Methods, devices, storage media and electronic equipment for detecting worker fatigue

Abstract

This application discloses a method, device, storage medium, and electronic device for detecting worker fatigue, relating to the field of occupational safety technology. The method includes: real-time acquisition of single-channel frontal EEG signals and head movement posture data from workers; establishing individualized EEG state baselines and resting motion baselines based on initial data acquired during resting or low-load states; extracting fatigue-sensitive features from the single-channel frontal EEG signals and extracting environmental motion features from the head movement posture data; constructing a fatigue judgment threshold based on the EEG state baseline and fatigue-sensitive features, and adaptively updating the threshold by introducing environmental motion features; combining the environmental motion features with the real-time fatigue-sensitive features and the updated threshold to complete a primary fatigue judgment; triggering a secondary confirmation through a deep analysis mode if the condition is suspected to be fatigued; and finally generating and outputting corresponding graded warning instructions by combining the fatigue state confirmation results with the head movement state analyzed by environmental motion features.

Description

Technical Field

[0001] This application relates to the field of work safety technology, and in particular to a method, device, storage medium and electronic equipment for detecting worker fatigue. Background Technology

[0002] In high-risk operations such as firefighting and emergency rescue, workers often need to work continuously for extended periods in complex environments characterized by high temperatures, high noise levels, and high workloads. This dual physical and mental strain easily leads to cognitive fatigue, decreased attention span, and slowed reaction times. Such fatigue significantly reduces the accuracy of workers' operations and their emergency response capabilities, thereby substantially increasing the probability of accidents and posing a serious threat to the safety of workers and the smooth progress of tasks. Therefore, real-time and accurate detection and early warning of worker fatigue has become an indispensable and crucial link in the safety assurance system for high-risk operations, and related fatigue detection technologies have become a key research focus in the fields of physiological signal monitoring and occupational safety technology.

[0003] In existing technologies, fatigue detection of workers largely relies on multi-channel EEG acquisition, complex sensor arrays, or behavioral visual analysis. Multi-channel EEG acquisition requires electrodes to be placed at multiple points on the head, along with complex signal processing units to extract features and determine the state. Behavioral visual analysis, on the other hand, requires high-definition cameras and image recognition algorithms to capture and analyze facial expressions, eye movements, and other behavioral characteristics of the worker. While these technologies can achieve fatigue detection to some extent, they generally suffer from complex equipment structures, high overall system power consumption, cumbersome multi-channel electrode placement, and stringent requirements for ambient lighting conditions, resulting in poor wearing comfort and insufficient adaptability to complex working environments. This makes it difficult to achieve long-term, stable fatigue monitoring in high-risk scenarios such as firefighting and emergency rescue.

[0004] Meanwhile, most existing fatigue detection technologies use fixed thresholds or single judgment models to identify fatigue states. These methods do not fully consider the individual physiological differences among workers and cannot adapt to the influence of dynamic factors such as environmental noise and changes in movement during operation, making them prone to false alarms or missed alarms in fatigue assessment. While some technical solutions attempt to optimize the judgment model, they lack effective environmental interference compensation and individual adaptation mechanisms. In the complex scenarios of actual high-risk operations, the stability and accuracy of the judgment results still fail to meet the needs of actual safety monitoring, and cannot provide reliable fatigue early warning support for workers. Summary of the Invention

[0005] In view of this, this application provides a method, device, storage medium and electronic device for detecting worker fatigue, which can achieve accurate detection of worker fatigue.

[0006] According to a first aspect of this application, a method for detecting worker fatigue is provided, the method being applied to a processing unit in a safety helmet, the safety helmet including a forehead single-channel EEG acquisition unit, a triaxial accelerometer, and the processing unit, the method comprising: The frontal single-channel EEG acquisition unit and the triaxial accelerometer collect the frontal single-channel EEG signal and head movement posture data of the operator in real time, and establish the operator's EEG status baseline and motion-resting baseline based on the initial acquisition data in the resting state or low-load operation state. Frequency band energy analysis was performed on the single-channel EEG signal in the frontal region to extract fatigue sensitivity features that characterize the alertness and cognitive load of the workers. At the same time, using the resting motion baseline as a reference, environmental motion features were extracted from the head motion posture data to help assess the quality of the EEG signal and the intensity of the work. A fatigue determination threshold is constructed based on the EEG state baseline and the fatigue sensitivity features, and the fatigue determination threshold is adaptively updated by incorporating the environmental motion features. Combining the environmental motion characteristics, the fatigue-sensitive features extracted in real time are compared with the adaptively updated fatigue judgment threshold to complete the initial fatigue judgment. For those whose judgment results indicate a suspected fatigue state, the deep analysis mode is triggered to confirm the fatigue state a second time. Based on the fatigue state confirmation result and the environmental motion characteristics, the head movement state of the worker is analyzed, and corresponding graded early warning instructions are generated.

[0007] According to a second aspect of this application, a device for detecting worker fatigue is provided. The device is applied to a processing unit within a safety helmet, the safety helmet including a forehead single-channel EEG acquisition unit, a triaxial accelerometer, and the processing unit. The device includes: The acquisition module is used to acquire the frontal single-channel EEG signal and head movement posture data of the operator in real time through the frontal single-channel EEG acquisition unit and the triaxial accelerometer, and to establish the operator's EEG state baseline and motor resting baseline based on the initial acquisition data in the resting state or low-load operation state. The extraction module is used to perform frequency band energy analysis on the frontal single-channel EEG signal, extract fatigue sensitivity features to characterize the alertness and cognitive load of the worker, and extract environmental motion features from the head motion posture data to help assess the quality of the EEG signal and the intensity of the work, based on the resting motion baseline. An update module is used to construct a fatigue judgment threshold based on the EEG state baseline and the fatigue sensitivity features, and to adaptively update the fatigue judgment threshold by incorporating the environmental motion features. The analysis module is used to combine the environmental motion features, compare the fatigue-sensitive features extracted in real time with the adaptively updated fatigue judgment threshold to complete the initial fatigue judgment. For the judgment result of suspected fatigue state, the deep analysis mode is triggered to confirm the fatigue state a second time. Based on the fatigue state confirmation result and the environmental motion features, the head movement state of the worker is analyzed to generate the corresponding graded early warning instruction.

[0008] According to a third aspect of this application, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for detecting worker fatigue.

[0009] According to a fourth aspect of this application, an electronic device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described method for detecting worker fatigue.

[0010] By utilizing the aforementioned technical solutions, the worker fatigue detection method, device, storage medium, and electronic equipment provided in this application integrate a frontal single-channel EEG acquisition unit with a triaxial accelerometer in a safety helmet. This single-channel EEG acquisition replaces existing multi-channel EEG, complex sensor arrays, and visual analysis methods, significantly simplifying the equipment structure and reducing system power consumption. Simultaneously, the helmet's design enhances worker comfort, making it suitable for complex and high-risk work environments such as firefighting and emergency rescue, enabling long-term, stable fatigue monitoring. Furthermore, by establishing individualized EEG baselines and resting-motor baselines, and adaptively updating fatigue judgment thresholds based on environmental motion characteristics, a multi-stage judgment mechanism combining primary fatigue judgment with secondary confirmation through in-depth analysis is constructed. This mechanism effectively adapts to the individual physiological differences of different workers and compensates for dynamic interference caused by environmental noise and changes in motion during operation. It can significantly reduce the probability of false alarms and false alarms in fatigue assessment, improve the stability and accuracy of assessment results, and generate graded early warning instructions by analyzing head movement status in conjunction with environmental motion characteristics. It can provide reliable fatigue status monitoring and early warning support for workers in high-risk work environments, effectively solving the problem that the assessment effect of existing technologies is not good and cannot meet the actual safety monitoring needs.

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This paper shows a structural schematic diagram of a smart safety helmet provided in an embodiment of this application; Figure 2 A schematic diagram illustrating the principle flow of a fatigue state detection and early warning method provided in an embodiment of this application is shown. Figure 3 A flowchart illustrating a method for detecting worker fatigue according to an embodiment of this application is shown. Figure 4 A flowchart illustrating a method for detecting worker fatigue according to another embodiment of this application is shown. Figure 5 A schematic diagram of the structure of a worker fatigue detection device provided in an embodiment of this application is shown. Detailed Implementation

[0013] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0014] In existing technologies, fatigue detection of workers largely relies on multi-channel EEG acquisition, complex sensor arrays, or behavioral visual analysis. Multi-channel EEG acquisition requires electrodes to be placed at multiple points on the head, along with complex signal processing units to extract features and determine the state. Behavioral visual analysis, on the other hand, requires high-definition cameras and image recognition algorithms to capture and analyze facial expressions, eye movements, and other behavioral characteristics of the worker. While these technologies can achieve fatigue detection to some extent, they generally suffer from complex equipment structures, high overall system power consumption, cumbersome multi-channel electrode placement, and stringent requirements for ambient lighting conditions, resulting in poor wearing comfort and insufficient adaptability to complex working environments. This makes it difficult to achieve long-term, stable fatigue monitoring in high-risk scenarios such as firefighting and emergency rescue.

[0015] Meanwhile, most existing fatigue detection technologies use fixed thresholds or single judgment models to identify fatigue states. These methods do not fully consider the individual physiological differences among workers and cannot adapt to the influence of dynamic factors such as environmental noise and changes in movement during operation, making them prone to false alarms or missed alarms in fatigue assessment. While some technical solutions attempt to optimize the judgment model, they lack effective environmental interference compensation and individual adaptation mechanisms. In the complex scenarios of actual high-risk operations, the stability and accuracy of the judgment results still fail to meet the needs of actual safety monitoring, and cannot provide reliable fatigue early warning support for workers.

[0016] In view of this, this application provides a safety helmet, such as Figure 1 As shown, the safety helmet uses the helmet body as its supporting base and integrates a frontal single-channel EEG acquisition unit, a three-axis accelerometer, and a processing unit. The frontal single-channel EEG acquisition unit consists of EEG acquisition electrodes, a reference electrode, and a ground electrode. The three-axis accelerometer, signal processing and control device (i.e., processing unit), power supply, warning output device, and signal acquisition unit are all located inside the helmet body. The components are connected by preset circuits. The power supply provides power to the signal processing and control device, signal acquisition unit, EEG acquisition unit, three-axis accelerometer, and warning output device. The frontal single-channel EEG acquisition unit and the three-axis accelerometer transmit the acquired frontal single-channel EEG signal and head movement posture data to the signal processing and control device, which completes the baseline establishment, feature extraction, threshold update, fatigue judgment, and secondary confirmation processes. Finally, the generated graded warning instructions are transmitted to the warning output device, realizing real-time monitoring and warning of the worker's fatigue status. At the same time, the sweatband fits the worker's forehead, improving wearing comfort and ensuring stable contact between the EEG acquisition electrodes and the skin.

[0017] Specifically, without altering the original protective structure and wearing method, the safety helmet can accommodate single-channel EEG acquisition electrodes in the forehead sweatband area inside the helmet, corresponding to the Fz point in the forehead region of the international 10–20 EEG system. A reference electrode is placed behind the user's ear, and a grounding electrode is placed in a non-signal-sensitive area inside the helmet. All acquisition electrodes utilize a flexible dry electrode structure, maintaining stable contact with the skin through the natural pressure generated when wearing the helmet. This eliminates the need for conductive gel, enabling continuous EEG signal acquisition and ensuring reliability and comfort under high temperature, high humidity, and prolonged wear conditions. The sampling rate of the EEG signal acquisition module can be set to 256 Hz to meet the sampling and analysis requirements of fatigue-related EEG frequency bands. The acquisition circuit employs a high-input-impedance bioelectric amplifier with a high common-mode rejection ratio to effectively suppress electromagnetic interference in the working environment and common-mode noise caused by human movement. The acquired raw EEG signals are first input to a preprocessing module, which includes bandpass filtering, notch filtering, and baseline drift suppression. The bandpass filter's passband range can be set from 0.5 to 40 Hz to remove low-frequency drift and high-frequency electromyography artifacts. Simultaneously, it performs notch filtering on 50 Hz power supply interference commonly found in industrial environments to improve the signal-to-noise ratio of EEG signals. Furthermore, the built-in triaxial accelerometer can synchronously acquire head motion data (Ax, Ay, Az) at a frequency of 50 Hz.

[0018] Accordingly, for embodiments of this disclosure, a method for detecting worker fatigue status applied to a processing unit in a safety helmet can be provided, such as... Figure 2 As shown, initial data of workers in resting or low-load working states can first be collected using a frontal single-channel EEG acquisition unit and a triaxial accelerometer to establish individualized EEG state baselines and resting motion baselines. Subsequently, frontal single-channel EEG signals and head movement posture data are acquired in real time, and frequency band energy analysis is performed on the EEG signals to extract [data / energy]. , Frequency band energy characteristics are used as fatigue-sensitive features, while the resultant displacement acceleration amplitude is extracted from head motion posture data based on the resting motion baseline as an environmental motion feature. Then, a low-power mode is entered, and a linear discriminant analysis model is constructed based on the EEG state baseline and fatigue-sensitive features to determine an individualized static fatigue judgment threshold characterizing fatigue risk. A dynamic reference background is then constructed based on real-time fatigue-sensitive features within a historical time window to determine the level of environmental noise fluctuations, and the fatigue judgment threshold is adaptively updated in conjunction with motion correction. Finally, the real-time extracted fatigue-sensitive features are compared with the adaptively updated threshold to complete the initial fatigue judgment. If a suspected fatigue state is determined, a high-precision mode is triggered. (i.e., deep analysis mode) constructs a single-class learning model based on samples of workers' normal states, integrates fatigue sensitivity features and environmental motion features to form a multi-dimensional feature vector, and further confirms whether the current features deviate from the individual's normal state space; at the same time, it independently conducts fall and disability monitoring based on environmental motion features. If an abnormal pulse (such as an impact) is detected and the worker remains stationary for a long time, it directly enters the early warning process; finally, it combines the fatigue state confirmation results of the high-precision mode with the fall and disability monitoring results to generate corresponding graded early warning instructions, and outputs prompt information through the early warning output device of the safety helmet or links with external safety systems to achieve long-term, stable and accurate fatigue state monitoring and safety early warning in high-risk work scenarios.

[0019] Specifically, such as Figure 3 As shown, the method for detecting worker fatigue may include the following steps: Step 310: Real-time acquisition of the frontal single-channel EEG signal and head movement posture data of the operator through the frontal single-channel EEG acquisition unit and the triaxial accelerometer, and establishment of the operator's EEG status baseline and motor resting baseline based on the initial acquisition data in the resting state or low-load operation state.

[0020] Among them, the workers are those wearing safety helmets integrated with a frontal single-channel EEG acquisition unit, a three-axis accelerometer, and a processing unit, engaged in high-risk operations such as firefighting and emergency rescue. They are prone to cognitive fatigue and inattention during the operation and require real-time monitoring and graded early warning of fatigue status. The frontal single-channel EEG signal is an electroencephalographic signal with only one channel acquired by the frontal single-channel EEG acquisition unit, used to characterize the worker's alertness and cognitive load status. The head movement posture data is acceleration data reflecting the three-dimensional movement state of the worker's head, acquired by the three-axis accelerometer, used to assess movement intensity and identify abnormal movement behavior. The EEG status baseline is a characteristic benchmark value obtained by statistically analyzing the worker's initial EEG signal in a resting or low-load working state, representing the level of EEG characteristics in a conscious and non-fatigued state. The motor resting baseline is a characteristic benchmark value obtained by statistically analyzing the worker's initial head movement posture data in a resting state, representing the level of motor characteristics when the head is in a static or stable state.

[0021] In this embodiment of the disclosure, the acquisition of two types of core data and the establishment of two types of individualized baselines can be completed simultaneously by integrating a frontal single-channel EEG acquisition unit and a triaxial accelerometer into a safety helmet: After the worker puts on the safety helmet, the initial frontal single-channel EEG signal and initial head movement posture data are first acquired when the worker is in a resting or low-load working state. Based on the initial data, the EEG state baseline representing the worker's awake state and the motion-resting baseline representing the head's stable state are statistically obtained. At the same time, the corresponding frontal single-channel EEG signal and head movement posture data are continuously acquired in real time during subsequent work, providing basic data and benchmark references for subsequent fatigue feature extraction, threshold construction and state determination, realizing a coherent data acquisition and baseline establishment process from initial benchmark calibration to real-time monitoring of the work process.

[0022] By adopting a single-channel frontal EEG acquisition method instead of a multi-channel acquisition method, the equipment structure and wearing process can be simplified, improving the wearing comfort and environmental adaptability of operators. At the same time, based on the initial acquisition data, an individualized EEG state baseline and motion-resting baseline can be established, which can provide a benchmark reference that fits the individual physiological characteristics for subsequent fatigue assessment. This can avoid the problem of insufficient adaptation due to individual differences caused by fixed thresholds, and lay the foundation for compensation for environmental motion interference and accurate assessment of fatigue state, ensuring the feasibility and accuracy of long-term and stable monitoring in high-risk work scenarios.

[0023] Step 320: Perform frequency band energy analysis on the single-channel EEG signal in the frontal region to extract fatigue sensitivity features used to characterize the alertness and cognitive load of the workers. At the same time, using the resting motion baseline as a reference, extract environmental motion features from the head motion posture data to help assess the quality of the EEG signal and the intensity of the work.

[0024] Among them, frequency band energy analysis is an analytical method that divides EEG signals into frequency bands and calculates the proportion or amplitude changes of signal energy in each frequency band, used to extract characteristic information reflecting the brain's alertness state; fatigue sensitivity characteristics are features extracted from EEG signals that are highly correlated with the alertness and cognitive load of workers, and can sensitively reflect changes in EEG activity under fatigue; the motion resting baseline is a benchmark value obtained based on the statistical data of head movement postures of workers in a resting state, representing the level of motion characteristics when the head is in a stable state; head movement posture data is acceleration data collected by a triaxial accelerometer that reflects the three-dimensional motion state of the worker's head, used to characterize the intensity and stability of head movement; environmental motion characteristics are characteristic indicators extracted from head movement posture data based on the motion resting baseline, used to assess the degree of motion interference to EEG signals and the intensity of work.

[0025] In this embodiment of the present disclosure, after data acquisition and baseline establishment are completed, frequency band energy analysis can be performed on the real-time acquired frontal single-channel EEG signal to extract fatigue sensitivity features that can effectively characterize the alertness and cognitive load of the worker. At the same time, with the pre-established motion resting baseline as a reference, the synchronously acquired head motion posture data is processed to extract environmental motion features used to assist in the assessment of EEG signal quality and work intensity. This achieves synchronous extraction of EEG features and motion features, providing multi-dimensional feature support for subsequent fatigue judgment threshold construction, interference compensation, and state judgment.

[0026] By extracting fatigue-sensitive features through frequency band energy analysis, subtle changes related to fatigue in EEG signals can be accurately captured, providing a core basis for fatigue state assessment. At the same time, by extracting environmental motion features based on the resting motion baseline, the degree of motion interference to EEG signals can be effectively assessed, providing a basis for interference compensation for subsequent adaptive threshold updates. It can also reflect the motion intensity of the worker, providing support for the identification of abnormal motion states. This achieves the synergistic extraction of core fatigue features and auxiliary environmental features, improving the anti-interference ability and accuracy of subsequent fatigue assessment.

[0027] Step 330: Construct a fatigue judgment threshold based on the EEG state baseline and fatigue sensitivity features, and introduce environmental motion features to adaptively update the fatigue judgment threshold.

[0028] The fatigue judgment threshold is a judgment standard determined based on the baseline of EEG status and fatigue sensitivity characteristics, used to distinguish whether the subject is in a fatigued state or not.

[0029] In this embodiment of the present disclosure, the baseline of the EEG state of the subject in a non-fatigue state can be determined first. At the same time, fatigue-sensitive features that can reflect the degree of fatigue can be extracted from the EEG signal. Based on the two, a fatigue determination threshold for distinguishing between fatigue and non-fatigue states can be constructed. In the fatigue determination process, environmental motion features in the environment in which the subject is located are introduced. The constructed fatigue determination threshold is dynamically and adaptively updated through environmental motion features to ensure that the fatigue determination threshold can adapt to changes in EEG state under different environmental motion conditions, thereby achieving more accurate fatigue determination.

[0030] This technical solution constructs a fatigue judgment threshold by using the baseline of EEG state as a reference and combining fatigue sensitivity features, which can ensure the basic accuracy of fatigue judgment. At the same time, it introduces environmental motion features to adaptively update the threshold, which can effectively solve the judgment deviation problem caused by EEG state fluctuations under different environmental motion conditions, improve the adaptability and accuracy of fatigue judgment, and more reliably realize the effective judgment of fatigue state, providing stable and accurate technical support for subsequent fatigue judgment-based applications.

[0031] Step 340: Combining environmental motion characteristics, the fatigue-sensitive features extracted in real time are compared with the adaptively updated fatigue judgment threshold to complete the initial fatigue judgment. For those whose judgment results are suspected of being in a fatigue state, the deep analysis mode is triggered to confirm the fatigue state a second time. Based on the fatigue state confirmation results and environmental motion characteristics, the head movement state of the worker is analyzed, and corresponding graded early warning instructions are generated.

[0032] Among them, suspected fatigue state is an intermediate state in the initial judgment where the features exceed the threshold but have not yet been finally confirmed, requiring further verification to avoid misjudgment; deep analysis mode is a high-precision verification process initiated for suspected fatigue state, which completes the final confirmation of fatigue state through more comprehensive feature analysis and model judgment; fatigue state confirmation result is the final fatigue state judgment conclusion obtained after verification by deep analysis mode, used to guide subsequent early warning decisions; head movement state is the head movement of the operator obtained based on environmental movement feature analysis, which can reflect the intensity of movement and whether there are abnormal states such as falls or prolonged stillness; graded early warning instructions are different levels of early warning signals generated based on fatigue state confirmation result and head movement state, used to differentiate and prompt work risks.

[0033] In this embodiment of the present disclosure, after feature extraction and threshold update are completed, the fatigue-sensitive features acquired in real time are compared with the adaptively updated fatigue judgment threshold in combination with environmental motion features to complete a rapid initial fatigue judgment. For cases judged as suspected fatigue states, a deep analysis mode is triggered to complete a secondary confirmation of the fatigue state through more comprehensive feature analysis and model verification. Then, the final fatigue state confirmation result is combined with the head movement state of the worker in the analysis of environmental motion features to finally generate a graded warning instruction that matches the current state, realizing a complete judgment process from rapid initial screening to accurate confirmation and then to graded warning.

[0034] This step, by combining initial fatigue assessment with environmental motion characteristics, enables rapid, low-power screening, ensuring monitoring efficiency while avoiding unnecessary computational consumption. The deep analysis mode, activated for suspected fatigue states, further improves the accuracy of fatigue assessment, effectively reducing false alarms and missed alarms. Simultaneously, by generating tiered early warning commands based on head movement status, it can simultaneously monitor abnormal movement hazards while identifying fatigue risks, providing workers with more comprehensive and accurate safety warning support and enhancing the reliability and practicality of monitoring in high-risk work scenarios.

[0035] In summary, the fatigue detection method for workers provided in this application integrates a single-channel forehead EEG acquisition unit with a triaxial accelerometer in a safety helmet. By replacing existing multi-channel EEG, complex sensor arrays, and visual analysis methods with single-channel EEG acquisition, the method significantly simplifies the equipment structure and reduces system power consumption. Furthermore, the helmet design enhances worker comfort, making it suitable for complex and high-risk work environments such as firefighting and emergency rescue, enabling long-term, stable fatigue monitoring. Additionally, by establishing individualized EEG baselines and resting motion baselines, and adaptively updating the fatigue threshold based on environmental motion characteristics, a multi-stage determination mechanism combining primary fatigue assessment with secondary confirmation through in-depth analysis is constructed. This mechanism effectively adapts to the individual physiological differences among workers and compensates for dynamic interference caused by environmental noise and changes in motion during operation. It can significantly reduce the probability of false alarms and false alarms in fatigue assessment, improve the stability and accuracy of assessment results, and generate graded early warning instructions by analyzing head movement status in conjunction with environmental motion characteristics. It can provide reliable fatigue status monitoring and early warning support for workers in high-risk work environments, effectively solving the problem that the assessment effect of existing technologies is not good and cannot meet the actual safety monitoring needs.

[0036] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, and to fully illustrate the implementation of this embodiment, this embodiment also provides another method for detecting worker fatigue, such as... Figure 4 As shown, the method includes: Step 410: Real-time acquisition of the frontal single-channel EEG signal and head movement posture data of the operator through the frontal single-channel EEG acquisition unit and the triaxial accelerometer, and establishment of the operator's EEG status baseline and motor resting baseline based on the initial acquisition data in the resting state or low-load operation state.

[0037] In this embodiment of the disclosure, when establishing the EEG state baseline and motor resting baseline of the worker based on the initial acquisition data under resting or low-load working conditions, the processing unit can first control the frontal single-channel EEG acquisition unit and the triaxial accelerometer to acquire the worker's initial frontal single-channel EEG signal and initial head movement posture data under the worker's resting or low-load working conditions for a preset duration. The acquired initial frontal single-channel EEG signal is then subjected to preprocessing operations such as bandpass filtering, notch filtering, and baseline drift suppression to completely remove noise and artifacts from the signal. Then, the preprocessed initial frontal single-channel EEG signal is subjected to short-time Fourier transform to obtain the power spectral density P(f) at each frequency point, and the absolute power of the Alpha band is calculated accordingly. Relative power characteristics ,in, The total power across the entire frequency band is the sum of the power spectral densities at all frequency points within the effective EEG signal's full frequency range. By performing frequency band energy statistics on this type of feature, a baseline for the EEG characteristics of the worker in a conscious state is determined, serving as the baseline for the EEG state. Simultaneously, the acceleration G is calculated by combining the initial head movement posture data. base This is used to determine the motion characteristic benchmark of the worker in the resting state, and to use it as the motion-resting baseline to complete the construction of two types of individualized baselines.

[0038] Accordingly, the implementation steps may include: controlling the frontal single-channel EEG acquisition unit and the triaxial accelerometer to acquire the initial frontal single-channel EEG signal and initial head movement posture data of the operator in a resting state or low-load operation state for a preset period of time; preprocessing the initial frontal single-channel EEG signal to remove noise and artifacts; performing frequency band energy statistics on the preprocessed initial frontal single-channel EEG signal to determine the EEG characteristic benchmark of the operator in a conscious state as the EEG state baseline; and performing modeling calculation on the initial head movement posture data to determine the movement characteristic benchmark of the operator in a resting state as the movement-resting baseline.

[0039] Among them, the preset duration refers to the time length set in advance for collecting initial data from workers in a resting or low-load working state to establish an effective individualized baseline, which is the time benchmark to ensure the statistical validity of the baseline; the initial frontal single-channel EEG signal is the raw EEG signal collected by the frontal single-channel EEG acquisition unit from workers in a resting or low-load working state, which is the basic data for constructing the EEG state baseline; the initial head movement posture data is the three-dimensional head movement acceleration data collected by the triaxial accelerometer from workers in a resting or low-load working state, which is the basic data for constructing the movement-resting baseline; preprocessing is the signal optimization processing operation performed on the raw initial EEG signal to eliminate various interferences in the signal and improve the quality and effectiveness of the EEG signal; frequency band energy statistics is a statistical analysis method for dividing the preprocessed EEG signal into frequency bands and calculating the signal energy of each frequency band, which is used to extract the EEG feature benchmark representing the awake state; and the modulus calculation is the modulus calculation operation performed on the three-dimensional acceleration data collected by the triaxial accelerometer to obtain a single feature value that can comprehensively reflect the head movement state.

[0040] This embodiment ensures the validity of baseline construction samples by controlling the acquisition unit to collect initial data under resting or low-load conditions within a preset time period. Multi-stage preprocessing of the initial EEG signals effectively removes noise and artifacts, improving the accuracy of EEG feature extraction. The EEG feature benchmark obtained by combining short-time Fourier transform and frequency band energy statistics can accurately characterize the worker's awake EEG state. The motion feature benchmark obtained by combining initial head movement posture data can comprehensively reflect the resting head movement state. Both baselines are constructed based on the worker's own physiological and motor characteristics, forming an individualized benchmark system that fits individual differences. This provides accurate and reliable reference for subsequent fatigue feature extraction, threshold construction, and adaptive updates, ensuring the accuracy and stability of subsequent fatigue state determination from the data source.

[0041] Step 420: Perform frequency band energy analysis on the single-channel EEG signal in the frontal region to extract fatigue sensitivity features that characterize the alertness and cognitive load of the workers. At the same time, using the resting baseline of motion as a reference, extract environmental motion features from the head motion posture data to help assess the quality of the EEG signal and the intensity of the work.

[0042] Fatigue sensitivity features can be divided into core fatigue sensitivity features and auxiliary fatigue sensitivity features. Core fatigue sensitivity features are the dominant frequency band energy features extracted from effective EEG signals that are highly correlated with the alertness and cognitive load of the worker, and are the core basis for characterizing fatigue state. Auxiliary fatigue sensitivity features are EEG frequency band energy features used in conjunction with core fatigue sensitivity features, which can supplement the reflection of the evolution trend of fatigue state and improve the comprehensiveness of fatigue feature characterization.

[0043] In this embodiment, the real-time acquired single-channel frontal EEG signal can first undergo bandpass filtering with a passband range of 0.5–40 Hz and notch filtering to remove 50 Hz power supply interference, eliminating low-frequency drift, high-frequency electromyography artifacts, and power supply interference to obtain the effective EEG signal. Then, a short-time Fourier transform is performed on the effective EEG signal to obtain the power spectral density P(f) at each frequency point. Based on this power spectral density, frequency band energy analysis is performed to calculate the absolute and relative power characteristics of the 8–13 Hz band. Frequency band energy characteristics are used as core fatigue-sensitive features, and 4–8Hz are extracted simultaneously. The energy characteristics of the frequency band are used as auxiliary fatigue-sensitive features to extract two types of fatigue-sensitive features. At the same time, the combined displacement amplitude is calculated from the real-time acquired three-axis head motion posture data (Ax, Ay, Az), and the calculation result is compared with the pre-established motion resting baseline G. base By comparison, environmental motion characteristics that can help assess the quality of EEG signals and the intensity of work can be obtained, and fatigue sensitivity characteristics and environmental motion characteristics can be extracted simultaneously.

[0044] Accordingly, the implementation steps may include: performing bandpass filtering and notch filtering on the real-time acquired frontal single-channel EEG signal to obtain an effective EEG signal; performing frequency band energy analysis on the effective EEG signal, extracting frequency band energy features as core fatigue-sensitive features, and simultaneously extracting frequency band energy features as auxiliary fatigue-sensitive features; calculating the combined displacement amplitude of the real-time acquired head motion posture data, and comparing the calculation results with the motion resting baseline to obtain environmental motion features used to assist in assessing the quality of EEG signals and work intensity.

[0045] By applying targeted bandpass and notch filtering to real-time EEG signals, the signal-to-noise ratio can be effectively improved, laying a data foundation for the accurate extraction of fatigue-sensitive features. Through frequency band energy analysis and differentiation of core and auxiliary fatigue-sensitive features, a comprehensive and accurate characterization of the alertness and cognitive load of workers can be achieved. Simultaneously, by calculating the combined displacement amplitude of head movement posture data and extracting environmental motion features in conjunction with the resting baseline, the degree of interference of EEG signals by motion artifacts can be effectively assessed, providing a basis for interference compensation in subsequent fatigue assessment. It can also accurately reflect the real-time work intensity of workers, achieving the synergistic extraction of core physiological features and auxiliary environmental features, significantly improving the anti-interference capability and accuracy of subsequent fatigue state assessment, and providing comprehensive and reliable multi-dimensional feature support for the entire fatigue detection process.

[0046] Step 430: Based on the fatigue sensitivity feature distribution of the EEG state baseline, construct a linear discriminant analysis model to determine the initial static fatigue judgment threshold.

[0047] Among them, the fatigue sensitivity feature distribution is the statistical distribution law of fatigue sensitivity features corresponding to the baseline EEG state of the worker in the feature space, reflecting the numerical range and statistical characteristics of features in the conscious and non-fatigue state; the linear discriminant analysis model is a classification model built based on the principle of linear discriminant analysis, which can effectively distinguish between fatigue and non-fatigue states by maximizing the inter-class distance of features in different states and minimizing the intra-class distance; the static fatigue judgment threshold is a fixed judgment criterion determined based on the EEG state baseline and the linear discriminant analysis model, which is the initial feature critical value for defining the transition of the worker from a conscious state to a fatigued state.

[0048] In the embodiments of this disclosure, when constructing a linear discriminant analysis model based on the fatigue sensitivity feature distribution of the EEG state baseline and determining the initial static fatigue judgment threshold, the continuous time window corresponding to the EEG state baseline is first... Frequency band relative power characteristics constitute a sober baseline feature sample set Calculate the feature mean vector of this sample set. and covariance matrix Then, the fatigue characteristic evolution law coefficient is introduced. Determine the feature drift amplitude Constructing a virtual fatigue class center μ fatigue Then, the optimal discriminant vector is calculated based on the principle of linear discriminant analysis. Projecting the conscious baseline sample onto the discriminant axis yields... This forms a projection sample set of conscious states, and the mean of this projection sample set is calculated. with standard deviation Finally, combined with the threshold adjustment coefficient λ, through The initial static fatigue judgment threshold was calculated, and the construction of the linear discriminant analysis model and the determination of the static threshold were completed.

[0049] This embodiment constructs a linear discriminant analysis model based on the fatigue sensitivity feature distribution of the EEG baseline. It can fully utilize the statistical regularities of the characteristics of workers in a conscious state. By constructing a virtual fatigue class center and calculating the optimal discriminant vector, it can effectively separate the feature space of fatigue and non-fatigue states. Combined with the statistical characteristics of the projected samples, a static fatigue judgment threshold is determined, allowing the threshold to accurately match the individual physiological characteristics of workers and form an initial fatigue judgment benchmark that fits the individual. This can effectively avoid the problem of poor adaptability of general fixed thresholds and lay an accurate and reliable foundation for subsequent adaptive updates of the threshold based on environmental factors. From the perspective of model construction and threshold setting, it ensures the individual adaptability and initial accuracy of fatigue judgment.

[0050] Step 440: Construct a dynamic reference background based on real-time fatigue sensitivity features within a historical time window. Statistically analyze the feature distribution of the dynamic reference background to obtain the noise fluctuation level of the current environment.

[0051] Among them, the historical time window is a pre-defined time range for selecting past fatigue-sensitive characteristic data during real-time fatigue monitoring, serving as the basis for defining the time dimension of the dynamic reference background; the dynamic reference background is a background system constructed based on real-time fatigue-sensitive characteristics within the historical time window, reflecting the statistical state of characteristic data in the current monitoring stage, used to characterize the characteristic fluctuation level caused by non-fatigue factors; the characteristic distribution is the statistical regularity and distribution state of real-time fatigue-sensitive characteristics within the dynamic reference background in the numerical space, serving as the core basis for analyzing environmental noise fluctuations; and the environmental noise fluctuation level is a quantitative indicator obtained through statistical analysis of the characteristic distribution of the dynamic reference background, characterizing the degree of influence of various non-physiological interference factors in the current working environment on fatigue-sensitive characteristics.

[0052] In the embodiments of this disclosure, when constructing a dynamic reference background based on real-time fatigue-sensitive features within a historical time window and analyzing the noise fluctuation level of the current environment, a sliding detection window containing detection units, protection units, and reference units can be constructed by sliding along the time axis on the continuously sampled fatigue-sensitive feature time series. The dynamic reference background is then constructed based on the real-time fatigue-sensitive features within the historical time window covered by the reference units. The set of reference units is denoted as... The historical feature samples included are real-time discriminant statistic sequences. and Then, statistical analysis is performed on the feature distribution within the dynamic reference background. The mean of the feature values ​​within the reference cell is used as the estimated background power value, and the result is obtained through the formula... Complete the calculation, where M is the reference element length and r i For the i-th historical feature sample within the reference cell, the background power estimate Z is... k This represents the current noise fluctuation level in the environment, fully realizing the construction of a dynamic reference background and the quantitative analysis of environmental noise fluctuation levels.

[0053] This embodiment constructs a dynamic reference background by defining historical time windows and building a sliding detection window containing multiple units. This enables the scientific selection and effective utilization of historical data on fatigue-sensitive features, avoids interference from short-term abrupt changes in features on background statistics, and ensures the authenticity and effectiveness of the background system. At the same time, by performing mean statistical analysis on the feature distribution of the dynamic reference background, the level of environmental noise fluctuation can be obtained, enabling precise quantification of the impact of non-physiological interference factors in the working environment. This provides an objective and reliable quantitative basis for subsequent adaptive compensation of fatigue judgment thresholds in conjunction with environmental factors, allowing threshold updates to accurately match the current environmental interference state and effectively improving the adaptability of subsequent fatigue judgments to environmental changes.

[0054] Step 450: Determine the environmental sensitivity adjustment coefficient based on the noise fluctuation level of the current environment, set the motion correction coefficient in combination with the environmental motion characteristics, and based on the constant false alarm rate control concept, perform dual weighted compensation on the environmental threshold corresponding to the dynamic reference background through the environmental sensitivity adjustment coefficient and the motion correction coefficient to obtain the environmental adaptive threshold that adapts to the current environment and motion state.

[0055] Among them, the environmental sensitivity adjustment coefficient is a quantized coefficient determined based on the noise fluctuation level of the current environment, used to adapt the environmental threshold adjustment under different noise intensities; the motion correction coefficient is a quantized coefficient set in combination with environmental motion characteristics, used to adjust the environmental threshold according to the intensity of the worker's head movement, and to counteract the interference of motion artifacts on fatigue judgment; the constant false alarm rate control idea is a control principle that keeps the probability of false alarms in fatigue judgment at a constant level under different environments and motion states. The core is to suppress the false alarm rate fluctuations caused by various interferences by dynamically adjusting the judgment threshold; the environmental threshold is the judgment limit corresponding to the dynamic reference background, reflecting the characteristic interference critical value under the current environmental noise fluctuation, and is the basis for constructing the environmental adaptive threshold; the dual weighted compensation is a weighted adjustment operation of the environmental threshold using both the environmental sensitivity adjustment coefficient and the motion correction coefficient to achieve dual compensation for environmental noise and motion interference; the environmental adaptive threshold is a dynamic judgment threshold obtained after dual weighted compensation that can adapt to the current environmental noise fluctuation and head movement state, and can be adjusted in real time with changes in environment and movement state.

[0056] In the embodiments of this disclosure, the environmental sensitivity adjustment coefficient can be determined based on the noise fluctuation level of the current environment, and the motion correction coefficient can be set in combination with the environmental motion characteristics. When performing double weighted compensation on the environmental threshold corresponding to the dynamic reference background based on the constant false alarm rate control concept to obtain the environmental adaptive threshold, the environmental sensitivity adjustment coefficient is first determined based on the quantified current environmental noise fluctuation level combined with the preset target false alarm probability. Simultaneously, a motion correction coefficient is set based on the resultant displacement acceleration amplitude corresponding to the real-time extracted environmental motion characteristics. Based on the concept of constant false alarm rate control, an offset compensation term δ is introduced, and the formula is used to calculate the offset compensation term δ. The environmental threshold corresponding to the dynamic reference background is weighted and calculated, where Z k G represents the estimated background power, which is the noise fluctuation level of the current environment. norm The resultant displacement acceleration amplitude of the head motion posture data is used. An environmental sensitivity adjustment coefficient is used to achieve weighted compensation for environmental noise fluctuations, and a motion correction coefficient is used to achieve weighted compensation for the head motion state. This completes a dual weighted compensation operation for the environmental threshold, resulting in the calculated value. This refers to the environmental adaptive threshold that adapts to the current environment and motion state.

[0057] This embodiment determines the environmental sensitivity adjustment coefficient based on the level of environmental noise fluctuation and sets the motion correction coefficient in combination with environmental motion characteristics. This enables targeted quantitative characterization of environmental and motion interference. Based on the constant false alarm rate control concept, the environmental threshold is double-weighted and compensated, allowing the environmental threshold to adapt to both the current environmental noise fluctuation state and head motion state. This effectively counteracts the dual interference of environmental noise and motion artifacts on fatigue judgment. The resulting adaptive environmental threshold can be dynamically adjusted according to changes in the working environment and motion state. This ensures the constant false alarm rate during fatigue judgment and improves the adaptability of the judgment threshold to complex working scenarios. It lays a dynamic benchmark that fits the real-time scenario for the subsequent construction of the final fatigue judgment threshold, and significantly reduces the risk of false alarms in fatigue judgment caused by environmental and motion factors.

[0058] Step 460: Compare the static fatigue judgment threshold with the environmental adaptive threshold, and take the maximum value as the adaptively updated fatigue judgment threshold.

[0059] In the embodiments of this disclosure, after constructing the static fatigue judgment threshold and the environmental adaptive threshold, the individualized initial static fatigue judgment threshold T determined based on the linear discriminant analysis model can be used. static The environmental adaptive threshold T, obtained by dual weighting compensation using environmental sensitivity adjustment coefficient and motion correction coefficient, is... env Numerical comparisons are performed using formulas. The maximum value between the two values ​​is calculated and used as the adaptively updated fatigue judgment threshold, thus completing the final construction of the fatigue judgment threshold and providing a judgment basis that combines individual physiological benchmarks and environmental movement adaptation for subsequent primary fatigue judgment.

[0060] This embodiment uses the maximum value between the static fatigue judgment threshold and the environmental adaptive threshold as the final fatigue judgment threshold. This ensures that the judgment threshold retains the basic judgment benchmark that fits the individual physiological characteristics of the operator, while also incorporating dynamic adjustment characteristics to adapt to the current environmental noise and motion state. This effectively avoids misjudgments caused by low thresholds due to environmental noise fluctuations or motion artifacts. At the same time, using the individual static threshold as the lower limit ensures the physiological rationality of fatigue judgment, so that the final judgment threshold can always maintain an accurate and reliable judgment standard in complex work scenarios. This provides dual protection for the accuracy and anti-interference of subsequent fatigue judgments from the threshold level.

[0061] Step 470: Combining environmental motion characteristics, the fatigue-sensitive features extracted in real time are compared with the adaptively updated fatigue judgment threshold to complete the initial fatigue judgment. For those whose judgment results indicate a suspected fatigue state, the deep analysis mode is triggered to confirm the fatigue state a second time. Based on the fatigue state confirmation results and environmental motion characteristics, the head movement state of the worker is analyzed, and corresponding graded early warning instructions are generated.

[0062] In this embodiment of the disclosure, when comparing the real-time extracted fatigue-sensitive features with the adaptively updated fatigue judgment threshold in conjunction with environmental motion characteristics to complete the initial fatigue judgment, the fatigue-sensitive features of the worker during the operation can be continuously extracted according to a preset time window, and integrated in chronological order to obtain a real-time fatigue-sensitive feature sequence. Then, each feature value in this sequence is compared with the adaptively updated fatigue judgment threshold T obtained by taking the maximum value of the static and environmental adaptive thresholds. dynamic Numerical comparisons are performed one by one to determine whether a single feature value exceeds the threshold, while simultaneously considering the resultant displacement acceleration amplitude G corresponding to the environmental motion characteristics. norm To determine whether the worker's head is in a state of non-violent impact, if any feature value in the real-time fatigue sensitivity feature sequence exceeds T... dynamic If the environmental motion characteristics indicate that the head is not in a state of severe impact, then the worker is directly determined to be in a state of suspected fatigue. If the characteristic value does not exceed the judgment threshold, or if the characteristic value exceeds the threshold but the head is not in a state of severe impact, then the worker is determined to be in a normal working state, thus completing the entire process of primary fatigue judgment.

[0063] Accordingly, the implementation steps may include: continuously extracting fatigue-sensitive features of workers during the work process according to a preset time window to obtain a real-time fatigue-sensitive feature sequence; comparing each feature value in the real-time fatigue-sensitive feature sequence with an adaptively updated fatigue judgment threshold one by one to determine whether the feature value exceeds the fatigue judgment threshold, and combining environmental motion features to determine whether the worker's head is in a non-violent impact state; if any feature value exceeds the fatigue judgment threshold and the worker's head is in a non-violent impact state, then the worker is determined to be in a suspected fatigue state; otherwise, the worker is determined to be in a normal working state.

[0064] The preset time window is a pre-defined time interval for continuously extracting fatigue-sensitive features from workers. It serves as a time benchmark to ensure real-time and continuous collection of fatigue features, providing a unified time dimension standard for subsequent sequence construction. The real-time fatigue-sensitive feature sequence is a sequence of feature values ​​formed by continuously extracting fatigue-sensitive features from workers during their work process according to the preset time window, dynamically reflecting the real-time changing trend of workers' fatigue features. The non-violent impact state is a stable motion state obtained based on environmental motion features, where the worker's head does not experience violent collisions or high-intensity movements. It is a prerequisite for eliminating motion artifacts and ensuring the effectiveness of fatigue determination. The suspected fatigue state is the determination result in the primary fatigue determination when the fatigue-sensitive feature value exceeds the determination threshold and the head is in a non-violent impact state. It is an intermediate fatigue state that needs further verification and has not yet formed a final fatigue determination conclusion. The normal working state is the determination result in the primary fatigue determination when the fatigue-sensitive feature value does not exceed the determination threshold, or when the feature value exceeds the threshold but the head is in a violent impact state. It represents that the worker currently has no obvious fatigue features or the determination is affected by motion interference and cannot confirm fatigue.

[0065] This embodiment extracts features and constructs a real-time fatigue-sensitive feature sequence according to a preset time window. It can dynamically and continuously capture changes in the fatigue characteristics of workers, providing comprehensive real-time data support for initial judgment. The feature sequence values ​​are compared one by one with the adaptively updated fatigue judgment threshold. Combined with the dual judgment logic of judging whether the head is in a non-violent impact state based on environmental motion characteristics, it can effectively eliminate the interference of violent motion artifacts on fatigue judgment and avoid misjudgment caused by motion interference. At the same time, by using the rule that any feature value exceeding the threshold and meeting the motion state condition is judged as a suspected fatigue state, potential fatigue risks can be quickly identified, achieving efficient initial screening of fatigue state. It can ensure the accuracy and anti-interference of the initial judgment, while also taking into account the real-time monitoring, providing accurate and reliable judgment basis for triggering subsequent deep analysis modes. At the same time, the low-complexity judgment logic can also ensure low power consumption operation of the system, adapting to the long-term monitoring needs of high-risk operation scenarios.

[0066] In this embodiment of the disclosure, when a suspected fatigue state is determined and a deep analysis mode is triggered for secondary confirmation of the fatigue state, the duration or number of consecutive determinations of the suspected fatigue state can be statistically analyzed in real time after the suspected fatigue state is determined. It can then be determined whether the statistical results meet the preset triggering conditions of the deep analysis mode. If they do, the deep analysis mode is immediately triggered, integrating the core fatigue sensitivity characteristics, auxiliary fatigue sensitivity characteristics, and environmental motion characteristics of the worker to construct a three-dimensional multidimensional feature vector. Then, based on the resultant displacement acceleration amplitude G corresponding to the environmental motion characteristics norm Conduct motion gating quality control, and include G norm With the preset threshold of intense exercise G th The system compares the current exercise intensity to determine if it exceeds a threshold. If the current exercise intensity exceeds a preset threshold for strenuous exercise, the system directly determines that the signal-to-noise ratio of the EEG signal is too low, suspends the current secondary fatigue status analysis, and maintains the previous assessment of the worker's status. If the current exercise intensity does not exceed the preset threshold for strenuous exercise, the constructed multidimensional feature vector is input into a pre-built single-class learning model based on samples of the worker's normal state. This model calculates the degree of deviation of the multidimensional feature vector from the individual's normal state space, and then compares the calculated deviation with the preset confirmation threshold. Based on the comparison result, the final fatigue status confirmation result of the worker is determined, completing the entire secondary fatigue status confirmation process.

[0067] Accordingly, the implementation steps may include: after determining that the worker is in a suspected fatigue state, firstly, statistically analyze the duration of the suspected fatigue state or the number of consecutive determinations to determine whether the triggering conditions for the deep analysis mode are met; if the triggering conditions are met, immediately trigger the deep analysis mode and construct a multi-dimensional feature vector containing fatigue sensitivity features and environmental motion features; perform motion gating quality control based on environmental motion features to determine whether the worker's current motion intensity exceeds a preset vigorous motion threshold; if the current motion intensity exceeds the preset vigorous motion threshold, determine that the EEG signal-to-noise ratio is too low, suspend the secondary confirmation analysis of the current fatigue state, and maintain the worker's state result determined at the previous moment; if the current motion intensity does not exceed the preset vigorous motion threshold, input the multi-dimensional feature vector into a pre-constructed single-class learning model based on the worker's normal state sample, calculate the deviation of the multi-dimensional feature vector from the individual's normal state space through the single-class learning model, compare the deviation with the preset confirmation threshold, and determine the worker's fatigue state confirmation result.

[0068] Among them, the multidimensional feature vector is a multidimensional feature set formed by integrating fatigue sensitivity features and environmental motion features. It can comprehensively reflect the physiological state of the worker and the state of the motion environment, providing a comprehensive input basis for the single-class learning model. Motion gating quality control refers to the signal quality control operation based on environmental motion features. It filters effective signals by judging motion intensity and eliminates low-quality EEG signals caused by strenuous exercise, ensuring the effectiveness of subsequent analysis. The preset strenuous exercise threshold is a quantitative critical value for defining whether the worker's head motion intensity belongs to strenuous exercise. It is the core reference standard for judging the quality of EEG signals in motion gating quality control. The single-class learning model is a machine learning model pre-constructed based on normal state samples of workers. It is trained only on normal state data and can accurately identify features. The deviation of the vector from the individual's normal state space; the individual's normal state space is a feature space constructed from multi-dimensional feature samples of the worker in normal state, representing the feature distribution range of the worker in non-fatigue state, and is the benchmark space for judging fatigue state; the deviation degree is a quantitative indicator of the distance between the multi-dimensional feature vector and the individual's normal state space, which can reflect the degree of difference between the worker's current state and normal state, and is the core basis for judging fatigue state; the preset confirmation threshold is a quantitative critical value for defining whether the deviation degree reaches the fatigue state, and is the judgment standard for secondary confirmation of fatigue state through a single-class learning model; the fatigue state confirmation result is the final fatigue state conclusion obtained after complete judgment by the deep analysis mode, which is different from the suspected fatigue state of the initial judgment, and is the core basis for generating graded early warning instructions.

[0069] This embodiment sets the trigger conditions for the deep analysis mode by statistically analyzing the duration of suspected fatigue states or the number of consecutive judgments. This effectively avoids invalid high-precision analysis caused by short-term feature fluctuations, balancing low power consumption and accuracy. It integrates fatigue-sensitive features and environmental motion features to construct a multi-dimensional feature vector, providing comprehensive feature data that combines physiological and environmental factors for secondary confirmation. Motion gating quality control based on environmental motion features effectively eliminates low-quality signals caused by strenuous exercise, ensuring the reliability of secondary confirmation from the data source. The effective multi-dimensional feature vector is input into a pre-constructed single-class learning model to calculate the deviation degree and compared with a preset confirmation threshold. This accurately quantifies the difference between the worker's current state and normal state, achieving high-precision confirmation of fatigue states and significantly reducing the false alarm probability of initial judgments. The entire deep analysis mode's judgment logic not only achieves accurate verification of suspected fatigue states but also achieves rational utilization of system resources through signal quality control and conditional triggering. It adapts to the long-term, high-precision fatigue monitoring needs in high-risk work scenarios, providing accurate and reliable judgment basis for the generation of subsequent graded early warning instructions.

[0070] In this embodiment of the present disclosure, when analyzing the head movement state of a worker based on fatigue state confirmation results and environmental motion characteristics to generate corresponding graded early warning instructions, the head movement state of the worker can be analyzed based on fatigue state confirmation results and environmental motion characteristics to generate corresponding graded early warning instructions. Specifically, based on the fatigue state confirmation results obtained from the deep analysis mode, combined with the degree of deviation of the multi-dimensional feature vector relative to the individual's normal state space, the degree of deviation is compared with multiple preset graded thresholds to classify and define the worker's fatigue level, determining whether it is mild fatigue, moderate fatigue, or severe fatigue. Simultaneously, based on the resultant displacement acceleration amplitude G corresponding to the environmental motion characteristics... norm Analyze the real-time head movement status of the workers and detect G. norm The system detects whether an abnormal pulse persists near zero for an extended period, thus identifying potential abnormal states such as falls or prolonged periods of stillness. If such an abnormal state is detected, the highest-level disability warning is generated directly. If no abnormal state is detected, a fatigue warning corresponding to the determined fatigue level is generated. Finally, the generated disability or fatigue warning is simultaneously sent to the safety helmet's warning output module, triggering the warning and ensuring the risk warning is effectively implemented.

[0071] Accordingly, the implementation steps may include: after confirming that the worker is in a fatigued state based on the fatigue state confirmation result, comparing the deviation of the multidimensional feature vector from the individual's normal state space with multiple preset grading thresholds, and determining the worker's fatigue level based on the threshold range in which the deviation falls, where fatigue levels include mild fatigue, moderate fatigue, and severe fatigue; analyzing the worker's real-time head movement state based on environmental motion characteristics to identify whether there are any abnormal states such as falls or prolonged stillness; if an abnormal state is identified, directly generating the highest level of disability warning instruction; if no abnormal state is identified, generating a fatigue warning instruction of the matching level based on the determined fatigue level; and simultaneously sending the generated disability warning instruction or fatigue warning instruction to the warning output module of the safety helmet to complete the output triggering of the warning instruction.

[0072] The fatigue level is a classification of the worker's fatigue level based on the fatigue status confirmation results, which can include mild fatigue, moderate fatigue, and severe fatigue; the abnormal status is an abnormal movement state of the worker's head, such as falling or prolonged stillness, identified based on environmental motion characteristics; the disability warning instruction is the highest-level warning signal generated for the identified abnormal status such as falling or prolonged stillness, used to alert the worker to the potential risk of disability; the fatigue warning instruction is a matching level warning signal generated according to the different fatigue levels of the worker, which can present differentiated warning prompts according to the degree of fatigue; the warning output module is a functional module in the safety helmet used to receive and output warning instructions, and is the terminal component for realizing the transmission of warning information and enabling the worker to perceive the risk.

[0073] This embodiment classifies the fatigue level of workers based on fatigue status confirmation results and generates matching fatigue warning instructions. This enables differentiated and precise fatigue risk alerts, allowing workers to intuitively perceive their fatigue level according to the warning level. It also identifies abnormal states such as falls and prolonged periods of inactivity based on environmental motion characteristics, directly triggering the highest-level disability warning instruction. This allows for rapid response to potential personal safety risks to workers, improving the timeliness and relevance of warnings. Simultaneously, various warning instructions are sent to the warning output module of the safety helmet to complete the output triggering. Utilizing the helmet's wearing characteristics, warning information can be transmitted in real-time at close range, ensuring workers quickly receive risk information. The entire graded warning logic integrates the dual determination of fatigue status and abnormal head movement states, enabling both tiered warnings of fatigue risk and emergency warnings for sudden personal safety risks during work. This forms a comprehensive, accurate, and efficient graded warning system, providing all-round safety warning support for workers in high-risk work scenarios and effectively reducing the risk of safety accidents caused by fatigue and sudden conditions.

[0074] In summary, the fatigue detection method for workers provided in this application integrates a single-channel forehead EEG acquisition unit with a triaxial accelerometer in a safety helmet. By replacing existing multi-channel EEG, complex sensor arrays, and visual analysis methods with single-channel EEG acquisition, the method significantly simplifies the equipment structure and reduces system power consumption. Furthermore, the helmet design enhances worker comfort, making it suitable for complex and high-risk work environments such as firefighting and emergency rescue, enabling long-term, stable fatigue monitoring. Additionally, by establishing individualized EEG baselines and resting motion baselines, and adaptively updating the fatigue threshold based on environmental motion characteristics, a multi-stage determination mechanism combining primary fatigue assessment with secondary confirmation through in-depth analysis is constructed. This mechanism effectively adapts to the individual physiological differences among workers and compensates for dynamic interference caused by environmental noise and changes in motion during operation. It can significantly reduce the probability of false alarms and false alarms in fatigue assessment, improve the stability and accuracy of assessment results, and generate graded early warning instructions by analyzing head movement status in conjunction with environmental motion characteristics. It can provide reliable fatigue status monitoring and early warning support for workers in high-risk work environments, effectively solving the problem that the assessment effect of existing technologies is not good and cannot meet the actual safety monitoring needs.

[0075] Furthermore, as Figure 3 and Figure 4 The specific implementation of the method shown in this embodiment provides a device for detecting worker fatigue, such as... Figure 5 As shown, the device includes: a data acquisition module 51, an extraction module 52, an update module 53, and an analysis module 54.

[0076] The acquisition module 51 can be used to acquire the frontal single-channel EEG signal and head movement posture data of the operator in real time through the frontal single-channel EEG acquisition unit and the triaxial accelerometer, and establish the operator's EEG status baseline and motor resting baseline based on the initial acquisition data in the resting state or low-load operation state. The extraction module 52 can be used to perform frequency band energy analysis on the single-channel EEG signal in the frontal region, extract fatigue sensitivity features to characterize the alertness and cognitive load of workers, and extract environmental motion features from head motion posture data to help assess the quality of EEG signals and the intensity of work, based on the resting motion baseline. The update module 53 can be used to construct a fatigue judgment threshold based on the EEG state baseline and fatigue sensitivity features, and to adaptively update the fatigue judgment threshold by introducing environmental motion features. The analysis module 54 can be used to combine environmental motion characteristics, compare the fatigue-sensitive features extracted in real time with the adaptively updated fatigue judgment threshold to complete the initial fatigue judgment. For the judgment result of suspected fatigue state, the deep analysis mode is triggered to confirm the fatigue state for a second time. Based on the fatigue state confirmation result and environmental motion characteristics, the head movement state of the worker is analyzed, and corresponding graded early warning instructions are generated.

[0077] In some embodiments of this application, the acquisition module 51 can be specifically used to control the frontal single-channel EEG acquisition unit and the triaxial accelerometer to acquire the initial frontal single-channel EEG signal and initial head movement posture data of the operator in a resting state or low-load operation state for a preset period of time; preprocess the initial frontal single-channel EEG signal to remove noise and artifacts; perform frequency band energy statistics on the preprocessed initial frontal single-channel EEG signal to determine the EEG characteristic benchmark of the operator in a conscious state as the EEG state baseline; and perform modeling calculation on the initial head movement posture data to determine the movement characteristic benchmark of the operator in a resting state as the movement-resting baseline.

[0078] In some embodiments of this application, fatigue sensitivity features include core fatigue sensitivity features and auxiliary fatigue sensitivity features; the extraction module 52 can specifically be used to perform bandpass filtering and notch filtering on the real-time acquired prefrontal single-channel EEG signal to obtain an effective EEG signal; and to perform frequency band energy analysis on the effective EEG signal to extract... Frequency band energy characteristics are used as core fatigue-sensitive features and are extracted simultaneously. Frequency band energy characteristics are used as auxiliary fatigue sensitivity features; the combined displacement amplitude is calculated from real-time acquired head movement posture data, and the calculation results are compared with the resting baseline of the movement to obtain environmental motion characteristics used to assist in the assessment of EEG signal quality and work intensity.

[0079] In some embodiments of this application, the update module 53 can be specifically used to construct a linear discriminant analysis model based on the fatigue sensitivity feature distribution of the EEG state baseline, and determine the initial static fatigue judgment threshold; construct a dynamic reference background based on the real-time fatigue sensitivity features within the historical time window, and obtain the noise fluctuation level of the current environment by statistically analyzing the feature distribution of the dynamic reference background; determine the environmental sensitivity adjustment coefficient according to the noise fluctuation level of the current environment, set the motion correction coefficient in combination with the environmental motion features, and, based on the constant false alarm rate control idea, perform double weighted compensation on the environmental threshold corresponding to the dynamic reference background through the environmental sensitivity adjustment coefficient and the motion correction coefficient to obtain the environmental adaptive threshold that adapts to the current environment and motion state; compare the static fatigue judgment threshold with the environmental adaptive threshold, and take the maximum value as the adaptively updated fatigue judgment threshold.

[0080] In some embodiments of this application, the analysis module 54 can be specifically used to continuously extract the fatigue-sensitive features of the worker during the operation according to a preset time window to obtain a real-time fatigue-sensitive feature sequence; compare each feature value in the real-time fatigue-sensitive feature sequence with the adaptively updated fatigue judgment threshold one by one to determine whether the feature value exceeds the fatigue judgment threshold, and combine the environmental motion features to determine whether the worker's head is in a non-violent impact state; if any feature value exceeds the fatigue judgment threshold and the worker's head is in a non-violent impact state, then the worker is determined to be in a suspected fatigue state; otherwise, the worker is determined to be in a normal working state.

[0081] In some embodiments of this application, the analysis module 54 can also be used to, after determining that the worker is in a suspected fatigue state, first statistically analyze the duration of the suspected fatigue state or the number of consecutive determinations to determine whether the triggering conditions for the deep analysis mode are met; if the triggering conditions are met, the deep analysis mode is immediately triggered to construct a multi-dimensional feature vector containing fatigue sensitivity features and environmental motion features; based on the environmental motion features, motion gating quality control is performed to determine whether the worker's current motion intensity exceeds a preset vigorous motion threshold; if the current motion intensity exceeds the preset vigorous motion threshold, the EEG signal-to-noise ratio is determined to be too low, the secondary confirmation analysis of the current fatigue state is suspended, and the worker's state result determined at the previous moment is maintained; if the current motion intensity does not exceed the preset vigorous motion threshold, the multi-dimensional feature vector is input into a single-class learning model pre-constructed based on the worker's normal state sample, and the deviation of the multi-dimensional feature vector from the individual's normal state space is calculated through the single-class learning model, and the deviation is compared with the preset confirmation threshold to determine the worker's fatigue state confirmation result.

[0082] In some embodiments of this application, the analysis module 54 can also be used to, after confirming that the worker is in a fatigued state based on the fatigue state confirmation result, compare the degree of deviation of the multidimensional feature vector relative to the individual's normal state space with multiple preset grading thresholds, and determine the worker's fatigue level according to the threshold range in which the degree of deviation is located. The fatigue level includes mild fatigue, moderate fatigue, and severe fatigue. Based on the environmental motion characteristics, the real-time head movement state of the worker is analyzed to identify whether there are abnormal states such as falls or prolonged stillness. If an abnormal state is identified, the highest level of disability warning instruction is directly generated. If no abnormal state is identified, a fatigue warning instruction of matching level is generated according to the determined fatigue level. The generated disability warning instruction or fatigue warning instruction is synchronously sent to the warning output module of the safety helmet to complete the output triggering of the warning instruction.

[0083] It should be noted that other corresponding descriptions of the functional units involved in the worker fatigue detection device provided in this embodiment can be found in [reference needed]. Figure 3 and Figure 4 The corresponding descriptions in [the document] will not be repeated here.

[0084] Based on the above, Figure 3 and Figure 4 Accordingly, this embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the above-described method. Figure 3 and Figure 4 The method for detecting worker fatigue is shown.

[0085] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause an electronic device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0086] Based on the above, Figure 3 and Figure 4 The method shown, and Figure 5 To achieve the above objectives, the present application also provides an electronic device, specifically a personal computer, tablet computer, server, or other network device, as shown in the virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to achieve the above-described objectives. Figure 3 and Figure 4 The method for detecting worker fatigue is shown.

[0087] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0088] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0089] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0090] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platform, or it can be implemented by hardware.

[0091] This invention integrates a single-channel forehead EEG acquisition unit with a triaxial accelerometer into a safety helmet. Replacing existing multi-channel EEG, complex sensor arrays, and visual analysis methods with single-channel EEG acquisition significantly simplifies the equipment structure and reduces system power consumption. Simultaneously, the helmet design improves worker comfort, making it suitable for complex and high-risk work environments such as firefighting and emergency rescue, enabling long-term, stable fatigue monitoring. Furthermore, by establishing individualized EEG baselines and resting-motor baselines, and adaptively updating fatigue judgment thresholds based on environmental motion characteristics, a multi-stage judgment mechanism combining primary fatigue judgment with secondary confirmation through in-depth analysis is constructed. This mechanism effectively adapts to the individual physiological differences of different workers and compensates for dynamic interference caused by environmental noise and changes in motion during operation. It significantly reduces the probability of false alarms and missed alarms in fatigue judgment, improving the stability and accuracy of the judgment results. Combined with analysis of head movement status based on environmental motion characteristics to generate tiered early warning commands, it provides reliable fatigue monitoring and early warning support for workers in high-risk work environments, effectively solving the problems of poor judgment performance and inability to meet actual safety monitoring needs in existing technologies.

[0092] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0093] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A workman fatigue state detection method characterized by comprising: The method is applied to a processing unit in a safety helmet, the safety helmet including a forehead single-channel EEG acquisition unit, a triaxial accelerometer, and the processing unit, the method comprising: The frontal single-channel EEG acquisition unit and the triaxial accelerometer collect the frontal single-channel EEG signal and head movement posture data of the operator in real time, and establish the operator's EEG status baseline and motion-resting baseline based on the initial acquisition data in the resting state or low-load operation state. Frequency band energy analysis was performed on the single-channel EEG signal in the frontal region to extract fatigue sensitivity features that characterize the alertness and cognitive load of the workers. At the same time, using the resting motion baseline as a reference, environmental motion features were extracted from the head motion posture data to help assess the quality of the EEG signal and the intensity of the work. A fatigue determination threshold is constructed based on the EEG state baseline and the fatigue sensitivity features, and the fatigue determination threshold is adaptively updated by incorporating the environmental motion features. Combining the environmental motion characteristics, the fatigue-sensitive features extracted in real time are compared with the adaptively updated fatigue judgment threshold to complete the initial fatigue judgment. For those whose judgment results indicate a suspected fatigue state, the deep analysis mode is triggered to confirm the fatigue state a second time. Based on the fatigue state confirmation result and the environmental motion characteristics, the head movement state of the worker is analyzed, and corresponding graded early warning instructions are generated.

2. The method of claim 1, wherein, Based on initial data collected during resting or low-load work conditions, the EEG baseline and motor-resting baseline of the workers are established, including: The frontal single-channel EEG acquisition unit and the triaxial accelerometer are controlled to acquire the initial frontal single-channel EEG signal and initial head movement posture data of the operator in a resting state or low-load operation state for a preset duration. The initial prefrontal single-channel EEG signal is preprocessed to remove noise and artifacts from the initial prefrontal single-channel EEG signal; Frequency band energy statistics were performed on the preprocessed initial frontal single-channel EEG signal to determine the EEG characteristic benchmark of the worker in a conscious state, which was used as the EEG state baseline. The initial head movement posture data is used to perform a modeling calculation to determine the motion characteristic benchmark of the worker in a resting state, which serves as the motion-resting baseline.

3. The method of claim 1, wherein, The fatigue sensitivity features include core fatigue sensitivity features and auxiliary fatigue sensitivity features; Frequency band energy analysis was performed on the single-channel frontal EEG signal to extract fatigue sensitivity features characterizing the worker's alertness and cognitive load. Simultaneously, using the resting motion baseline as a reference, environmental motion features were extracted from the head movement posture data to aid in assessing EEG signal quality and workload, including: Bandpass filtering and notch filtering were performed on the real-time acquired single-channel frontal EEG signal to obtain an effective EEG signal. performing frequency band energy analysis on the effective brain electrical signals to extract a frequency band energy feature as the core fatigue sensitive feature a frequency band energy feature as the auxiliary fatigue sensitive feature The combined displacement amplitude of the real-time collected head movement posture data is calculated, and the calculation result is compared with the resting motion baseline to obtain environmental motion characteristics used to assist in the assessment of EEG signal quality and work intensity.

4. The method of claim 1, wherein, A fatigue determination threshold is constructed based on the aforementioned EEG baseline and the aforementioned fatigue sensitivity features, and the fatigue determination threshold is adaptively updated by incorporating the aforementioned environmental motion features, including: Based on the fatigue sensitivity feature distribution of the aforementioned EEG state baseline, a linear discriminant analysis model is constructed to determine the initial static fatigue judgment threshold. A dynamic reference background is constructed based on real-time fatigue sensitivity features within a historical time window. The noise fluctuation level of the current environment is obtained by statistically analyzing the feature distribution of the dynamic reference background. The environmental sensitivity adjustment coefficient is determined based on the noise fluctuation level of the current environment, and the motion correction coefficient is set in combination with the environmental motion characteristics. Based on the constant false alarm rate control concept, the environmental threshold corresponding to the dynamic reference background is double-weighted and compensated by the environmental sensitivity adjustment coefficient and the motion correction coefficient to obtain the environmental adaptive threshold that adapts to the current environment and motion state. The static fatigue determination threshold is compared with the environmental adaptive threshold, and the maximum value is taken as the adaptively updated fatigue determination threshold.

5. The method of claim 1, wherein, Combining the aforementioned environmental motion characteristics, the fatigue-sensitive features extracted in real time are compared with the adaptively updated fatigue determination threshold to complete the initial fatigue determination, including: The fatigue sensitivity features of the workers during the work process are continuously extracted according to a preset time window to obtain a real-time fatigue sensitivity feature sequence. Each feature value in the real-time fatigue-sensitive feature sequence is compared with the adaptively updated fatigue judgment threshold to determine whether the feature value exceeds the fatigue judgment threshold, and the environmental motion features are combined to determine whether the worker's head is in a non-violent impact state. If any of the aforementioned feature values ​​exceeds the fatigue determination threshold and the worker's head is not in a state of severe impact, then the worker is determined to be in a state of suspected fatigue; otherwise, the worker is determined to be in a normal working state.

6. The method of claim 5, wherein, For cases where the assessment result indicates a suspected fatigue state, a deep analysis mode is triggered to conduct a secondary confirmation of the fatigue state, including: After determining that the worker is in a suspected state of fatigue, the duration of the suspected fatigue state or the number of consecutive determinations is statistically analyzed to determine whether the triggering conditions of the deep analysis mode are met. If the triggering condition is met, the deep analysis mode is immediately triggered to construct a multi-dimensional feature vector containing the fatigue sensitivity feature and the environmental motion feature; Based on the environmental motion characteristics, motion gating quality control is performed to determine whether the current motion intensity of the worker exceeds a preset vigorous motion threshold. If the current exercise intensity exceeds the preset intense exercise threshold, the EEG signal-to-noise ratio is determined to be too low, the secondary confirmation analysis of the current fatigue state is suspended, and the worker's state result determined at the previous moment is maintained. If the current exercise intensity does not exceed the preset intense exercise threshold, the multidimensional feature vector is input into a pre-constructed single-class learning model based on the normal state samples of the worker. The single-class learning model calculates the degree of deviation of the multidimensional feature vector relative to the individual's normal state space. The degree of deviation is compared with a preset confirmation threshold to determine the worker's fatigue state confirmation result.

7. The method of claim 6, wherein, Based on the fatigue status confirmation results and the environmental motion characteristics, the operator's head movement status is analyzed to generate corresponding graded early warning instructions, including: After confirming that the worker is in a fatigued state based on the fatigue state confirmation result, the degree of deviation of the multidimensional feature vector relative to the individual's normal state space is compared with multiple preset grading thresholds. Based on the threshold range in which the degree of deviation is located, the fatigue level of the worker is determined. The fatigue level includes mild fatigue, moderate fatigue, and severe fatigue. Based on the environmental motion characteristics, the real-time head movement status of the workers is analyzed to identify any abnormal states such as falls or prolonged stillness. If the abnormal state is detected, the highest level of disability warning instruction will be generated directly. If the abnormal state is not identified, a fatigue warning instruction matching the determined fatigue level is generated. The generated disability warning command or fatigue warning command is simultaneously sent to the warning output module of the safety helmet to complete the output triggering of the warning command.

8. A device for detecting worker fatigue, characterized in that, The device is used as a processing unit in a safety helmet, the safety helmet including a forehead single-channel EEG acquisition unit, a triaxial accelerometer, and the processing unit, the device comprising: The acquisition module is used to acquire the frontal single-channel EEG signal and head movement posture data of the operator in real time through the frontal single-channel EEG acquisition unit and the triaxial accelerometer, and to establish the operator's EEG state baseline and motor resting baseline based on the initial acquisition data in the resting state or low-load operation state. The extraction module is used to perform frequency band energy analysis on the frontal single-channel EEG signal, extract fatigue sensitivity features to characterize the alertness and cognitive load of the worker, and extract environmental motion features from the head motion posture data to help assess the quality of the EEG signal and the intensity of the work, based on the resting motion baseline. An update module is used to construct a fatigue judgment threshold based on the EEG state baseline and the fatigue sensitivity features, and to adaptively update the fatigue judgment threshold by incorporating the environmental motion features. The analysis module is used to combine the environmental motion features, compare the fatigue-sensitive features extracted in real time with the adaptively updated fatigue judgment threshold to complete the initial fatigue judgment. For the judgment result of suspected fatigue state, the deep analysis mode is triggered to confirm the fatigue state a second time. Based on the fatigue state confirmation result and the environmental motion features, the head movement state of the worker is analyzed to generate the corresponding graded early warning instruction.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

10. An electronic device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

Images