Fatigue monitoring method and system for electric power scene

Abstract

The invention discloses a fatigue monitoring method and system for an electric power scene, and the method comprises the steps: collecting an electroencephalogram signal of an electric power worker, and obtaining original electroencephalogram signal data; preprocessing the original electroencephalogram signal data to obtain a purified electroencephalogram signal; calculating a power spectrum density and a frequency band energy ratio based on the purified electroencephalogram signal; based on the power spectrum density and the frequency band energy ratio, generating a preliminary fatigue index and a fatigue probability, and performing weighted fusion on the preliminary fatigue index and the fatigue probability to obtain a fatigue level; and when the fatigue level is greater than a preset value, triggering acousto-optic early warning. Therefore, the calculation complexity during state updating is reduced, the time consumption of single filtering is reduced to be within an equipment sampling interval, the real-time response capability of fatigue monitoring is ensured, and the timeliness requirement of an electric power scene on online monitoring is met.

Description

Technical Field

[0001] This invention relates to the field of electroencephalogram (EEG) signal processing and safety monitoring technology, and more specifically, to a fatigue monitoring method and system for power scenarios. Background Technology

[0002] Power industry operations (such as substation operation and maintenance, and line inspection) are characterized by high intensity and high risk. Worker fatigue directly affects operational accuracy and can easily lead to safety accidents. Existing fatigue monitoring technologies have the following limitations: (1) Video-based monitoring (such as blink frequency analysis) is susceptible to light and obstruction, and has low reliability in outdoor or complex power scenarios; (2) Monitoring based on a single physiological signal (such as heart rate, skin conductance) indirectly reflects fatigue status, but its sensitivity is insufficient; (3) Existing EEG signal monitoring devices mostly use a single filtering method, which makes it difficult to effectively separate EEG characteristic frequency bands (Delta, Theta, Alpha, Beta, Gamma waves), and the feature fusion method is simple, resulting in large errors in fatigue level assessment.

[0003] Therefore, there is an urgent need for a fatigue monitoring device that can adapt to power scenarios, has strong anti-interference capabilities, and high evaluation accuracy. Summary of the Invention

[0004] According to the present invention, a fatigue monitoring method and system for power scenarios are provided to solve the technical problems of existing fatigue monitoring devices having weak anti-interference ability, insufficient extraction of EEG features, and low accuracy of fatigue assessment in power scenarios.

[0005] According to a first aspect of the present invention, a fatigue monitoring method for a power scenario is provided, comprising: Collect EEG signals from power workers to obtain raw EEG signal data; The raw EEG signal data is preprocessed to obtain purified EEG signals; Based on the purified EEG signal, the power spectral density and the frequency band energy ratio are calculated. Based on the power spectral density and the frequency band energy ratio, a preliminary fatigue index and fatigue probability are generated. The preliminary fatigue index and fatigue probability are then weighted and fused to obtain the fatigue level.

[0006] Optionally, the raw EEG signal data is preprocessed to obtain purified EEG signals, including: The original EEG signal was input into a fourth-order Butterworth bandpass filter to separate the signals of each frequency band. The Kalman filter is used to denoise the signals in each frequency band, and the denoising result is expressed as follows: Equations of state: The observation equation is , in Let the state transition matrix be a first-order identity matrix. Let the observation matrix be set to 1; Based on the denoising results, an adaptive factor is introduced. Dynamically optimize the Kalman filter gain to obtain an adaptive noise reduction signal by introducing an adaptive factor. The calculation method is as follows: , in, The adjustment coefficient is 0.1-0.5. Predict the state at time k; Principal component analysis (PCA) was performed on the adaptive noise-reduced signal to remove artifacts from electrooculography (EOG) and head movement (HMM) signals, resulting in purified EEG signals.

[0007] Optionally, the raw EEG signal is input into a 4th-order Butterworth bandpass filter to separate the signals in each frequency band, including: The transfer function of the Butterworth filter is determined as follows: in, This is the cutoff angular frequency for the corresponding frequency band. It is a complex frequency variable.

[0008] The signals in each frequency band include Delta waves (0.5-4Hz), which are associated with deep sleep and increase with accumulated fatigue; Theta waves (4-8Hz), which are associated with drowsiness and increase significantly with fatigue; Alpha waves (8-13Hz), which are associated with wakefulness and relaxation and decrease with fatigue; Beta waves (13-30Hz), which are associated with alertness and decrease with fatigue; and Gamma waves (30-50Hz), which are associated with cognitive activity and fluctuate with fatigue.

[0009] Optionally, principal component analysis (PCA) is performed on the adaptively denoised signal to reduce dimensionality and remove electrooculography (EOG) and head movement artifacts, resulting in a purified EEG signal, including: Principal component analysis (PCA) was performed on the adaptive noise-reduced signal to remove electrooculography (EOG) and head movement artifacts through dimensionality reduction. The specific calculation method is as follows: Center the filtered signal matrix: Calculate the covariance matrix and obtain the eigenvalues. right Eigenvalue decomposition, reconstructing the signal from principal components whose cumulative contribution rate is greater than a predetermined value: The effective brainwave components are preserved, resulting in purified brainwave signals.

[0010] Optionally, based on the purified EEG signal, the power spectral density and band energy ratio are calculated, including: The power spectral density (PSD) of the purified EEG signals in each frequency band is calculated using the following formula: , in, Sampling frequency, For window length, For the number of windows, For the first Fourier transform results for each window; Calculate the total energy of the frequency band based on the power spectral density (PSD). , Calculate the frequency band energy ratio: .

[0011] Optionally, based on the power spectral density and the frequency band energy ratio, a preliminary fatigue index and fatigue probability are generated. The preliminary fatigue index and fatigue probability are then weighted and fused to obtain the fatigue level, including: Based on the frequency band energy ratio, a preliminary fatigue index is generated, expressed as: , in, These are the weighting coefficients; The power spectral density (PSD) and the band energy ratio are input into a convolutional neural network (CNN) for training, which outputs the fatigue probability. ; The fatigue level is obtained by weighting and combining the preliminary fatigue index with the fatigue probability, using the following formula: in,( ), fusion coefficient , .

[0012] Optionally, the convolutional neural network (CNN) includes a 10-dimensional feature vector in the input layer, 32 5×1 kernels in the first convolutional layer, a max pooling layer with a stride of 2, 64 3×1 kernels in the second convolutional layer, 32 neurons in the fully connected layer, ReLU activation, and a Sigmoid output layer.

[0013] According to another aspect of the present invention, a fatigue monitoring system for power scenarios is also provided, comprising: The EEG signal acquisition module is used to collect EEG signals from power workers and obtain raw EEG signal data. The EEG signal purification module is used to preprocess the original EEG signal data to obtain purified EEG signals. The energy ratio calculation module is used to calculate the power spectral density and the frequency band energy ratio based on the purified EEG signal. A fatigue level module is obtained, which is used to generate a preliminary fatigue index and fatigue probability based on power spectral density and frequency band energy ratio. The preliminary fatigue index and fatigue probability are weighted and fused to obtain the fatigue level.

[0014] According to another aspect of the invention, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method as described in any one of the claims. According to another aspect of the present invention, an electronic device is also provided, comprising: The computer-readable storage medium; and one or more processors for executing a program in the computer-readable storage medium.

[0015] Therefore, by employing a dynamic Kalman gain update mechanism, signal tracking accuracy is significantly improved compared to traditional fixed-parameter filtering methods in non-steady-state noise environments. This is particularly beneficial in highly dynamic power scenarios, effectively avoiding estimation lag caused by sudden signal changes. When the observed noise suddenly increases (such as transient electromagnetic interference from equipment), the system automatically reduces the gain weight, weakening the impact of abnormal noise on the estimated value. Simultaneously, it updates the covariance matrix to ensure filtering stability, overcoming the overfitting or underconvergence defects of conventional Kalman filtering due to its fixed parameters. This reduces the computational complexity of state updates, lowering the single-cycle filtering time to within the equipment sampling interval, ensuring real-time response capability for fatigue monitoring and meeting the timeliness requirements of online monitoring in power scenarios. Attached Figure Description

[0016] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures: Figure 1 This is a flowchart illustrating a fatigue monitoring method for a power scenario as described in this embodiment. Figure 2This is a schematic diagram of a fatigue monitoring system for a power scenario as described in this embodiment. Detailed Implementation

[0017] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0018] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.

[0019] According to a first aspect of the invention, a method 100 is provided, with reference to Figure 1 As shown, the method 100 includes: S101: Collect EEG signals from power workers to obtain raw EEG signal data; S102: Preprocess the raw EEG signal data to obtain purified EEG signals; S103: Based on the purified EEG signal, calculate the power spectral density and the frequency band energy ratio; S104: Based on the power spectral density and the frequency band energy ratio, a preliminary fatigue index and fatigue probability are generated. The preliminary fatigue index and fatigue probability are weighted and fused to obtain the fatigue level.

[0020] Specifically, a wearable dry electrode headband is used as the data acquisition terminal, eliminating the need for conductive gel and adapting to mobile scenarios in power operations. The acquisition channels include the forehead (FP1, FP2), temporal lobe (T3, T4), and occipital lobe (O1, O2) regions, covering areas with active EEG signals; the sampling frequency is set to 500Hz to balance signal integrity and data volume.

[0021] The raw EEG signals are purified, including: A fourth-order Butterworth bandpass filter is used to separate five characteristic frequency bands by setting different cutoff frequencies: - Delta waves: 0.5-4Hz (related to deep sleep, increases with accumulated fatigue); - Theta wave: 4-8Hz (related to drowsiness, significantly enhanced when fatigued); - Alpha waves: 8-13Hz (related to a state of wakefulness and relaxation, weakens when fatigued); - Beta wave: 13-30Hz (related to alertness, weakens when fatigued); - Gamma wave: 30-50Hz (related to cognitive activity, fluctuates during fatigue).

[0022] The filter transfer function is optimized for each frequency band. Taking the Theta wave (4-8Hz) as an example, the cutoff angular frequency is... and This ensures the accuracy of frequency band separation.

[0023] Kalman filtering is used to process the bandpass signal, suppressing residual noise such as power frequency interference (50Hz) and electromyography artifacts. Its core equation is: Equations of state: in Let be the state transition matrix, and set it as a first-order identity matrix; Observation equation: in Let the observation matrix be set to 1; Process noise and observation noise The variances are respectively set as and The optimal estimate is calculated recursively. .

[0024] To address signal abrupt changes in power scenarios (such as electromagnetic interference from equipment), an adaptive factor is introduced. Dynamically adjust Kalman gain: in When the signal changes abruptly (residual) When (increases), Increase to enhance the robustness of the filter.

[0025] Artifact Removal Unit: Principal Component Analysis (PCA) is used to remove artifacts such as electrooculography (EOG) and head movements. For the filtered signal matrix ( For the number of channels, (For the number of sampling points) Centralized: Calculate the covariance matrix right Eigenvalue decomposition, reconstructing signals from principal components with a cumulative contribution rate > 95%: Preserve effective brainwave components.

[0026] Fatigue-related features were extracted from the purified signal, including: Welch's method is used to estimate the PSD of each frequency band. The signal is divided into a window of length 2s (1000 points) (with 50% overlap). A Hanning window is added to reduce spectral leakage. The formula is as follows: in , , For the number of windows, For the first The Fourier transform result of the window.

[0027] Calculate the ratio of total energy and fatigue-sensitive energy for each frequency band: Total energy of the frequency band (Theta wave and Alpha wave): Similarly, calculate ; Energy ratio: Fatigue levels are output through multi-model fusion, including: Bandwidth energy ratio model: weight coefficients trained based on historical data Calculate the preliminary fatigue index and normalize it to 0-1: Deep learning model: Employs 1D-CNN to process multi-domain features (PSD value + energy ratio), with the following structure: Input layer (10-dimensional features) → Convolutional layer 1 (32 x 5×1 kernels, ReLU) → Pooling layer (2×1, stride 2) → Convolutional layer 2 (64 x 3×1 kernels, ReLU) → Pooling layer (2×1, stride 2) → Fully connected layer (32 neurons, ReLU) → Output layer (Sigmoid, output...) ).

[0028] Fusion Unit: Weighted fusion of the outputs of two models to obtain fatigue levels 1-5. Level 1 represents no fatigue, while Level 5 represents extreme fatigue.

[0029] An audible and visual alarm device is used. When the fatigue level output by the fusion assessment module is ≥3 (moderate or above fatigue), a continuous buzzer (80dB) and red light flashing (frequency 2Hz) are triggered to prompt the operator to stop the operation.

[0030] Optionally, the raw EEG signal data is preprocessed to obtain purified EEG signals, including: The original EEG signal was input into a fourth-order Butterworth bandpass filter to separate the signals of each frequency band. The Kalman filter is used to denoise the signals in each frequency band, and the denoising result is expressed as follows: Equations of state: The observation equation is , in Let the state transition matrix be a first-order identity matrix. Let the observation matrix be set to 1; Based on the denoising results, an adaptive factor is introduced. Dynamically optimize the Kalman filter gain to obtain an adaptive noise reduction signal by introducing an adaptive factor. The calculation method is as follows: , in, The adjustment coefficient is 0.1-0.5. Predict the state at time k; Principal component analysis (PCA) was performed on the adaptive noise-reduced signal to remove artifacts from electrooculography (EOG) and head movement (HMM) signals, resulting in purified EEG signals.

[0031] Optionally, the raw EEG signal is input into a 4th-order Butterworth bandpass filter to separate the signals in each frequency band, including: The transfer function of the Butterworth filter is determined as follows: in, This is the cutoff angular frequency for the corresponding frequency band. It is a complex frequency variable.

[0032] The signals in each frequency band include Delta waves (0.5-4Hz), which are associated with deep sleep and increase with accumulated fatigue; Theta waves (4-8Hz), which are associated with drowsiness and increase significantly with fatigue; Alpha waves (8-13Hz), which are associated with wakefulness and relaxation and decrease with fatigue; Beta waves (13-30Hz), which are associated with alertness and decrease with fatigue; and Gamma waves (30-50Hz), which are associated with cognitive activity and fluctuate with fatigue.

[0033] Optionally, principal component analysis (PCA) is performed on the adaptively denoised signal to reduce dimensionality and remove electrooculography (EOG) and head movement artifacts, resulting in a purified EEG signal, including: Principal component analysis (PCA) was performed on the adaptive noise-reduced signal to remove electrooculography (EOG) and head movement artifacts through dimensionality reduction. The specific calculation method is as follows: Center the filtered signal matrix: Calculate the covariance matrix and obtain the eigenvalues. right Eigenvalue decomposition, reconstructing the signal from principal components whose cumulative contribution rate is greater than a predetermined value: The effective brainwave components are preserved, resulting in purified brainwave signals.

[0034] Optionally, based on the purified EEG signal, the power spectral density and band energy ratio are calculated, including: The power spectral density (PSD) of the purified EEG signals in each frequency band is calculated using the following formula: , in, Sampling frequency, For window length, For the number of windows, For the first Fourier transform results for each window; Calculate the total energy of the frequency band based on the power spectral density (PSD). , Calculate the frequency band energy ratio: .

[0035] Optionally, based on the power spectral density and the frequency band energy ratio, a preliminary fatigue index and fatigue probability are generated. The preliminary fatigue index and fatigue probability are then weighted and fused to obtain the fatigue level, including: Based on the frequency band energy ratio, a preliminary fatigue index is generated, expressed as: , in, These are the weighting coefficients; The power spectral density (PSD) and the band energy ratio are input into a convolutional neural network (CNN) for training, which outputs the fatigue probability. ; The fatigue level is obtained by weighting and combining the preliminary fatigue index with the fatigue probability, using the following formula: in,( ), fusion coefficient , .

[0036] Optionally, the convolutional neural network (CNN) includes a 10-dimensional feature vector in the input layer, 32 5×1 kernels in the first convolutional layer, a max pooling layer with a stride of 2, 64 3×1 kernels in the second convolutional layer, 32 neurons in the fully connected layer, ReLU activation, and a Sigmoid output layer.

[0037] Therefore, by employing a dynamic Kalman gain update mechanism, signal tracking accuracy is significantly improved compared to traditional fixed-parameter filtering methods in non-steady-state noise environments. This is particularly beneficial in high-dynamic power scenarios, effectively avoiding estimation lag caused by abrupt signal changes. When the observed noise suddenly increases (such as transient electromagnetic interference from equipment), the system automatically reduces the gain weight to weaken the impact of abnormal noise on the estimated value. Simultaneously, it updates the covariance matrix to ensure filtering stability, overcoming the overfitting or underconvergence defects of conventional Kalman filtering due to its fixed parameters. This reduces the computational complexity of state updates, lowering the single-cycle filtering time to within the equipment sampling interval, ensuring real-time response capability for fatigue monitoring and meeting the timeliness requirements of online monitoring in power scenarios.

[0038] According to another aspect of the present invention, a fatigue monitoring system 200 for power scenarios is also provided, with reference to... Figure 2 As shown, the system 200 includes: The EEG signal acquisition module 210 is used to collect the EEG signals of power workers and obtain raw EEG signal data. The EEG signal purification module 220 is used to preprocess the original EEG signal data to obtain purified EEG signals. The energy ratio calculation module 230 is used to calculate the power spectral density and the frequency band energy ratio based on the purified EEG signal. A fatigue level module 240 is obtained, which is used to generate a preliminary fatigue index and fatigue probability based on power spectral density and frequency band energy ratio, and to obtain a fatigue level by weighted fusion of the preliminary fatigue index and fatigue probability.

[0039] The fatigue monitoring system 200 for a power scenario according to an embodiment of the present invention corresponds to the fatigue monitoring method 100 for a power scenario according to another embodiment of the present invention, and will not be described again here.

[0040] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0041] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0042] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0043] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0044] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0045] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method of fatigue monitoring of an electrical power scenario, characterized by, The method comprises the following steps: Collecting the brain electrical signals of power operation personnel to obtain raw brain electrical signal data; Pretreating the raw brain electrical signal data to obtain purified brain electrical signals; Based on the purified brain electrical signals, calculating the power spectral density and the frequency band energy ratio; Based on the power spectral density and the frequency band energy ratio, generating a preliminary fatigue index and a fatigue probability, and fusing the preliminary fatigue index and the fatigue probability by weighting to obtain a fatigue grade.

2. The fatigue monitoring method of claim 1, wherein, The pretreatment of the raw brain electrical signal data to obtain the purified brain electrical signals comprises the following steps: Inputting the raw brain electrical signals into a fourth-order Butterworth band-pass filter to separate each frequency band signal; Carrying out denoising on the each frequency band signal by Kalman filtering to obtain a denoising result, and the expression is: State equation: The observation equation is , wherein is a state transition matrix, let be a 1st order identity matrix, is an observation matrix, let be 1; According to the denoising result, an adaptive factor is introduced The dynamic optimization Kalman filter gain is obtained, the adaptive denoising signal is obtained, and the adaptive factor is introduced The calculation method is as follows: ， wherein, is a tuning factor between 0.1 and 0.5, is a predicted state at time k. Carrying out principal component analysis (PCA) on the adaptive denoising signal to reduce the dimension and remove the electrooculogram artifact and head movement artifact to obtain the purified brain electrical signals.

3. The fatigue monitoring method of claim 2, wherein, The inputting of the raw brain electrical signals into the fourth-order Butterworth band-pass filter to separate the each frequency band signal comprises the following steps: Determining the transfer function of the Butterworth filter as: wherein is the cut-off corner frequency for the corresponding frequency band, is the complex frequency variable; Wherein, the each frequency band signal comprises a Delta wave 0.5-4 Hz, which is related to deep sleep and is enhanced when fatigue accumulates; a Theta wave 4-8 Hz, which is related to drowsiness and is significantly enhanced when fatigue; an Alpha wave 8-13 Hz, which is related to a wakeful and relaxed state and is weakened when fatigue; a Beta wave 13-30 Hz, which is related to an alert state and is weakened when fatigue; and a Gamma wave 30-50 Hz, which is related to cognitive activity and fluctuates when fatigue.

4. The fatigue monitoring method of claim 2, wherein, The principal component analysis (PCA) on the adaptive denoising signal to reduce the dimension and remove the electrooculogram artifact and head movement artifact to obtain the purified brain electrical signals comprises the following steps: The principal component analysis (PCA) on the adaptive denoising signal to reduce the dimension and remove the electrooculogram artifact and head movement artifact is specifically calculated as follows: Centralizing the filtered signal matrix: computing the covariance matrix, obtaining eigenvalues To Eigenvalue decomposition, reconstructing the signal using principal components whose cumulative contribution rate is greater than a predetermined value: Retaining the effective brain electrical components to obtain the purified brain electrical signals.

5. The fatigue monitoring method of claim 2, wherein, Based on the purified brain electrical signals, calculating the power spectral density and the frequency band energy ratio comprises the following steps: Calculating the power spectral density (PSD) of the purified each frequency band brain electrical signal, and the calculation formula is: ， wherein, is a sampling frequency, is a window length, is a number of windows, is a Fourier transform result of the th window; Based on the power spectral density (PSD), calculating the total energy of the frequency band ， Calculating the frequency band energy ratio: 。 6. The fatigue monitoring method of claim 5, wherein, Based on the power spectral density and the frequency band energy ratio, generating a preliminary fatigue index and a fatigue probability, and fusing the preliminary fatigue index and the fatigue probability by weighting to obtain a fatigue grade comprises the following steps: Based on the frequency band energy ratio, generating a preliminary fatigue index, and the expression is: ， wherein are weight coefficients; inputting the power spectral density PSD and the band energy ratio into a convolutional neural network (CNN) for training, and outputting a fatigue probability ; Fusing the preliminary fatigue index and the fatigue probability by weighting to obtain a fatigue grade, and the formula is: wherein (A) is a compound of formula (I) ), fusion coefficient , .

7. The fatigue monitoring method according to claim 6, wherein The convolutional neural network (CNN) comprises an input layer 10-dimensional feature vector, a first convolutional layer 32 5×1 kernels, a maximum pooling layer step 2, a second convolutional layer 64 3×1 kernels, a fully connected layer 32 neurons, a ReLU activation and a Sigmoid output layer.

8. A fatigue monitoring system for an electrical power scenario, characterized by The method comprises the following steps: An electroencephalogram signal acquisition module is configured to collect the brain electrical signals of power operation personnel to obtain raw brain electrical signal data; A purified electroencephalogram signal module is configured to preprocess the raw electroencephalogram signal data to obtain a purified electroencephalogram signal; A calculation energy ratio module is configured to calculate a power spectral density and a frequency band energy ratio based on the purified electroencephalogram signal; A fatigue grade obtaining module is configured to generate a preliminary fatigue index and a fatigue probability based on the power spectral density and the frequency band energy ratio, and to obtain a fatigue grade by weightedly fusing the preliminary fatigue index and the fatigue probability.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program, when executed by a processor, implements the steps of the method of any one of claims 1-7.

10. An electronic device, comprising: The program comprises: The computer readable storage medium of claim 9; and One or more processors configured to execute the program in the computer readable storage medium.

Images