Fatigue detection method and device, electronic equipment and computer readable storage medium

Abstract

The invention discloses a fatigue detection method. The method comprises the following steps: detecting electroencephalogram signals corresponding to a plurality of signal acquisition positions on a head; determining two-dimensional positions of the plurality of signal acquisition positions according to the relative position relationship of the plurality of signal acquisition positions; corresponding theta wave energy characteristics, delta wave energy characteristics and alpha wave periodic energy characteristics are determined according to the electroencephalogram signals corresponding to the signal acquisition positions on the two-dimensional position, and a pixel value corresponding to each signal acquisition position on the two-dimensional position is determined; wherein the pixel values reflect the theta wave energy characteristic, the delta wave energy characteristic and the alpha wave periodic energy characteristic of the electroencephalogram signal corresponding to each signal acquisition position on the two-dimensional position; determining an electroencephalogram information image according to the pixel value corresponding to each signal acquisition position on the two-dimensional position; the electroencephalogram information image is used for fatigue judgment. The method can improve the accuracy of fatigue detection. The invention further discloses a fatigue detection device, electronic equipment and a computer readable storage medium.

Description

Technical Field

[0001] The present invention relates to the field of biomedical engineering technology, and in particular to a fatigue detection method, a fatigue detection device, an electronic device and a computer-readable storage medium. Background Art

[0002] Every year, there are many traffic accidents caused by fatigue, especially in the field of aviation, where pilot fatigue can seriously threaten flight safety. EEG signals can directly reflect human neural activity, so they are considered to be able to be used for fatigue detection. EEG waves usually include the following main bands, namely delta waves (0.5-4Hz), theta waves (4-8Hz), alpha waves (8-14Hz) and beta waves (14-30Hz). These bands are associated with different brain states, for example, alpha waves are associated with a state of relaxation, while theta waves are associated with fatigue or deep thinking.

[0003] In the prior art, neural activity is visualized, and then combined with the image mining capabilities of deep learning, the image is recognized and fatigue detection is performed. When constructing an image, the prior art uses the energy features of theta wave, the energy features of the alpha wave, and the energy features of the beta wave, or calculates the energy features of theta wave, the energy features of the alpha wave, and the energy features of the beta wave, and uses (α+θ) / β, (α+θ) / (α+β), and (θ / β) to construct an image, but this construction method of the prior art is still not accurate enough for fatigue detection. Summary of the invention

[0004] The present invention proposes a fatigue detection method to solve the above problem.

[0005] In a first aspect, an embodiment of the present invention discloses a fatigue detection method, comprising:

[0006] Detecting EEG signals corresponding to multiple signal collection positions on the head;

[0007] Determining two-dimensional positions of the plurality of signal collection positions according to a relative positional relationship of the plurality of signal collection positions;

[0008] Determine the corresponding θ wave energy feature, δ wave energy feature and α wave periodic energy feature according to the EEG signals corresponding to the multiple signal collection positions on the two-dimensional position, and determine the pixel value corresponding to each signal collection position on the two-dimensional position; wherein the pixel value reflects the magnitude of the θ wave energy feature, δ wave energy feature and α wave periodic energy feature of the EEG signal corresponding to each signal collection position on the two-dimensional position;

[0009] The electroencephalogram information image is determined according to the pixel value corresponding to each signal collection position in the two-dimensional position; the electroencephalogram information image is used for fatigue judgment.

[0010] By adopting the above technical scheme, the fatigue detection method of the present invention determines the pixel value corresponding to each signal acquisition position in a two-dimensional position by specifically selecting the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal, and then obtains the EEG information image used for fatigue judgment. Compared with other EEG characteristics, the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics can better reflect the user's fatigue condition, and therefore can improve the accuracy of fatigue judgment.

[0011] According to another specific embodiment of the present invention, the alpha wave periodic energy characteristics of the EEG signal are determined by:

[0012] Determine the logarithmic power spectrum of the EEG signal;

[0013] According to the logarithmic power spectrum, nonlinear fitting is performed using the following formula 1 to obtain the corresponding fitting curve:

[0014] L(f)=ab×log(P(f)) (Formula 1)

[0015] Wherein, L(f) is the fitting curve, P(f) is the logarithmic power spectrum, a is the offset parameter of the fitting curve, b is the scaling parameter of the fitting curve, a and b are determined by least squares fitting, f is the frequency of the EEG signal, and the range of f is 2-5 Hz and 14-30 Hz;

[0016] The α wave frequency band energy is determined based on the logarithmic power spectrum and the fitting curve, and the α wave periodic energy characteristic is the α wave frequency band energy.

[0017] According to another specific embodiment of the present invention, determining the alpha wave band energy according to the logarithmic power spectrum and the fitting curve specifically includes:

[0018] The alpha wave spectrum periodic component is determined based on the logarithmic power spectrum and the fitting curve using the following formula 2:

[0019] P′(f)=P(f)-L(f) (Formula 2)

[0020] Among them, P′(f) is the periodic component of the α wave spectrum, L(f) is the fitting curve, P(f) is the logarithmic power spectrum, and f is the frequency of the EEG signal;

[0021] The α wave spectrum periodic energy component is determined by the following formula 3 based on the α wave spectrum periodic component:

[0022] P″(f)=10×10 P′(f) (Formula 3)

[0023] Among them, P′(f) is the periodic component of the α wave spectrum, P″(f) is the periodic energy component of the α wave spectrum, and f is the frequency of the EEG signal;

[0024] The α wave frequency band energy is determined based on the α wave spectrum period energy component.

[0025] According to another specific embodiment of the present invention, determining the two-dimensional positions of the plurality of signal collection positions according to the relative positional relationship of the plurality of signal collection positions specifically includes:

[0026] Determine an initial two-dimensional position according to the relative position relationship of multiple signal collection positions; the initial two-dimensional position includes a signal collection position and a blank position, and the blank position is a position where no EEG signal is detected;

[0027] In the initial two-dimensional position, the blank position on the left side of the middle column of the initial two-dimensional position is filled with the type of the position on the right side of the blank position, the blank position on the right side of the middle column of the initial two-dimensional position is filled with the type of the position on the left side of the blank position, and the blank position in the middle column of the initial two-dimensional position is filled with the type of the position located above and below the blank position that is relatively closer to the center of the middle column, to obtain the two-dimensional position.

[0028] According to another specific embodiment of the present invention, the plurality of signal acquisition positions are determined by the positions of electrodes placed in the 10-20 system electrode placement method.

[0029] According to another specific embodiment of the present invention, the pixel value corresponding to each signal acquisition position in the two-dimensional position is determined by the following method:

[0030] Determine the corresponding red channel EEG energy feature, green channel EEG energy feature and blue channel EEG energy feature according to the θ wave energy feature, δ wave energy feature and α wave periodic energy feature corresponding to each signal collection position on the two-dimensional position;

[0031] Normalizing the red channel EEG energy feature, the green channel EEG energy feature, and the blue channel EEG energy feature respectively to obtain the red channel pixel value, the green channel pixel value, and the blue channel pixel value;

[0032] Determine a pixel value based on a red channel pixel value, a green channel pixel value, and a blue channel pixel value.

[0033] According to another specific embodiment of the present invention, the red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are normalized respectively to obtain the red channel pixel value, the green channel pixel value and the blue channel pixel value, specifically including:

[0034] The red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are respectively normalized by the following formula 4, formula 5 and formula 6 to obtain the red channel pixel value, the green channel pixel value and the blue channel pixel value:

[0035] r′ i =((r i -min(r)) / (max(r)-min(r)) (Formula 4)

[0036] g′ i =((g i -min(g)) / (max(g)-min(g)) (Formula 5)

[0037] b′ i =((b i -min(b)) / (max(b)-min(b)) (Formula 6)

[0038] Among them, r i represents the red channel EEG energy feature corresponding to the EEG signal measured at the i-th signal acquisition position, r′ i Indicates r i Normalized red channel EEG energy feature, g i represents the green channel EEG energy feature corresponding to the EEG signal measured at the i-th signal acquisition position, g′ i Indicates g i Normalized green channel EEG energy characteristics, b i represents the blue channel EEG energy feature corresponding to the EEG signal measured at the i-th signal acquisition position, b′ i Indicates b i The normalized blue channel EEG energy feature, max(r) represents the maximum value in the set of multiple known red channel EEG energy features, min(r) represents the minimum value in the set of multiple known red channel EEG energy features, max(g) represents the maximum value in the set of multiple known green channel EEG energy features, min(g) represents the minimum value in the set of multiple known green channel EEG energy features, max(b) represents the maximum value in the set of multiple known blue channel EEG energy features, and min(b) represents the minimum value in the set of multiple known blue channel EEG energy features.

[0039] According to another specific embodiment of the present invention, a set of multiple known red channel EEG energy features includes red channel EEG energy features corresponding to multiple awake EEG signals and red channel EEG energy features corresponding to multiple fatigue EEG signals, a set of multiple known green channel EEG energy features includes green channel EEG energy features corresponding to multiple awake EEG signals and green channel EEG energy features corresponding to multiple fatigue EEG signals, and a set of multiple known blue channel EEG energy features includes blue channel EEG energy features corresponding to multiple awake EEG signals and blue channel EEG energy features corresponding to multiple fatigue EEG signals.

[0040] According to another specific embodiment of the present invention, the red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are normalized respectively to obtain the red channel pixel value, the green channel pixel value and the blue channel pixel value, further comprising:

[0041] Based on the normalized results obtained by formula 4, formula 5 and formula 6, the following formula 7, formula 8 and formula

[0042] Formula 9 is normalized again to obtain the red channel pixel value, the green channel pixel value and the blue channel pixel value:

[0043]

[0044]

[0045]

[0046] Among them, r″ i represents the final normalized red channel EEG energy feature corresponding to the i-th signal acquisition position, g″ i represents the final normalized green channel EEG energy feature corresponding to the i-th signal acquisition position, b″ i represents the final normalized blue channel EEG energy feature corresponding to the i-th signal acquisition position, r′ i represents the red channel EEG energy feature normalized by Formula 4 corresponding to the i-th signal acquisition position, g′ i represents the green channel EEG energy feature normalized by Formula 5 corresponding to the i-th signal acquisition position, b′ i represents the blue channel EEG energy feature normalized by formula 6 corresponding to the i-th signal acquisition position, and k is 7-13.

[0047] According to another specific embodiment of the present invention, the corresponding red channel EEG energy features, green channel EEG energy features and blue channel EEG energy features are determined according to the θ wave energy features, δ wave energy features and α wave periodic energy features corresponding to each signal acquisition position on the two-dimensional position, specifically including:

[0048] The red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are respectively determined to be one of the following three types: theta wave energy feature, delta wave energy feature, alpha wave periodic energy feature, and the red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are different from each other.

[0049] According to another specific embodiment of the present invention, it also includes:

[0050] The EEG information image is subjected to image recognition through the fatigue detection model to determine whether the EEG information image represents a wakeful state or a fatigued state.

[0051] According to another specific embodiment of the present invention, detecting the EEG signals corresponding to the multiple signal collection positions on the head specifically includes:

[0052] Detecting the original EEG signals corresponding to multiple signal collection positions on the head;

[0053] The original EEG signal is preprocessed to obtain an EEG signal; wherein the preprocessing includes one or more of the following methods: filtering, removing power frequency interference, bad channel detection and interpolation, independent component analysis, re-referencing, data cleaning and screening.

[0054] In a second aspect, an embodiment of the present invention discloses a fatigue detection device, comprising:

[0055] An EEG signal detection module is used to detect EEG signals corresponding to multiple signal collection positions on the head;

[0056] A two-dimensional position determination module, used to determine the two-dimensional positions of multiple signal collection positions according to the relative position relationship of the multiple signal collection positions;

[0057] A pixel value determination module is used to determine the corresponding θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics according to the EEG signals corresponding to the plurality of signal acquisition positions on the two-dimensional position, and determine the pixel value corresponding to each signal acquisition position on the two-dimensional position; wherein the pixel value reflects the magnitude of the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal corresponding to each signal acquisition position on the two-dimensional position;

[0058] The EEG information image determination module is used to determine the EEG information image according to the pixel value corresponding to each signal collection position in the two-dimensional position; the EEG information image is used for fatigue judgment.

[0059] By adopting the above technical scheme, the fatigue detection device of the present invention determines the pixel value corresponding to each signal collection position in a two-dimensional position by specifically selecting the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal, and then obtains the EEG information image used for fatigue judgment. Compared with other EEG characteristics, the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics can better reflect the user's fatigue condition, and therefore can improve the accuracy of fatigue judgment.

[0060] In a third aspect, an embodiment of the present invention discloses an electronic device, including a processor and a memory, wherein the memory stores at least one instruction, and when the at least one instruction is executed by the processor, the fatigue detection method in any embodiment of the first aspect is implemented.

[0061] By adopting the above technical scheme, the electronic device determines the pixel value corresponding to each signal collection position in the two-dimensional position by specifically selecting the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal, and then obtains the EEG information image used for fatigue judgment. Compared with other EEG characteristics, the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics can better reflect the user's fatigue status, and therefore can improve the accuracy of fatigue judgment.

[0062] In a fourth aspect, an embodiment of the present invention discloses a computer-readable storage medium, in which at least one instruction is stored. When the at least one instruction is executed, the fatigue detection method in any embodiment of the first aspect is implemented.

[0063] By adopting the above technical scheme, a computer-readable storage medium determines the pixel value corresponding to each signal acquisition position in a two-dimensional position by specifically selecting the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal, thereby obtaining an EEG information image for fatigue judgment. Compared with other EEG characteristics, the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics can better reflect the user's fatigue condition, and therefore can improve the accuracy of fatigue judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A schematic diagram showing the process of the fatigue detection method in an embodiment of the present invention Figure 1 ;

[0065] Figure 2 A schematic diagram showing the process of the fatigue detection method in an embodiment of the present invention Figure 2 ;

[0066] Figure 3 A schematic diagram showing the process of the fatigue detection method in an embodiment of the present invention Figure 3 ;

[0067] Figure 4 A schematic diagram showing the process of the fatigue detection method in an embodiment of the present invention Figure 4 ;

[0068] Figure 5 A schematic diagram showing the process of the fatigue detection method in an embodiment of the present invention Figure 5 ;

[0069] Figure 6 A schematic diagram showing the structure of a fatigue detection device in an embodiment of the present invention is shown;

[0070] Figure 7 A schematic diagram showing the structure of an electronic device in an embodiment of the present invention is shown;

[0071] Figure 8 A schematic diagram showing the positions of electrodes placed by the 10-20 system electrode placement method in an embodiment of the present invention;

[0072] Fig. 9 A schematic diagram showing a logarithmic power spectrum and a fitting curve in an embodiment of the present invention;

[0073] Fig.10 A schematic diagram showing the periodic energy component of the alpha wave spectrum in an embodiment of the present invention;

[0074] Fig.11 The EEG information images of fatigue and wakefulness in nine tests corresponding to normalization by Formula 4, Formula 5 and Formula 6 in an embodiment of the present invention are shown;

[0075] Fig.12 Schematic diagram showing the EEG information images of fatigue and wakefulness in nine tests corresponding to normalization by Formula 4, Formula 5, Formula 6, Formula 7, Formula 8 and Formula 9 in an embodiment of the present invention Figure 1 ;

[0076] Fig.13 Schematic diagram showing the EEG information images of fatigue and wakefulness in nine tests corresponding to normalization by Formula 4, Formula 5, Formula 6, Formula 7, Formula 8 and Formula 9 in an embodiment of the present invention Figure 2 ;

[0077] Fig.14 Schematic diagram showing the EEG information images of fatigue and wakefulness in nine tests corresponding to normalization by Formula 4, Formula 5, Formula 6, Formula 7, Formula 8 and Formula 9 in an embodiment of the present invention Figure 3 ;

[0078] Fig.15 The confusion matrix corresponding to the test set used to test the fatigue detection model in the embodiment of the present invention is shown;

[0079] Fig.16 A statistical schematic diagram showing the occurrence frequency of different EEG features in an embodiment of the present invention. DETAILED DESCRIPTION

[0080] The following specific embodiments illustrate the implementation of the present invention, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Although the description of the present invention will be introduced in conjunction with the preferred embodiment, this does not mean that the features of this invention are limited to this implementation. On the contrary, the purpose of introducing the invention in conjunction with the implementation is to cover other options or modifications that may extend based on the claims of the present invention. In order to provide a deep understanding of the present invention, the following description will include many specific details. The present invention can also be implemented without using these details. In addition, in order to avoid confusion or blurring the focus of the present invention, some specific details will be omitted in the description. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0081] It should be noted that in this specification, similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0082] The terms “first”, “second”, etc. are only used for distinguishing descriptions and should not be understood as indicating or implying relative importance.

[0083] In the description of this embodiment, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this embodiment can be understood according to specific circumstances.

[0084] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0085] According to research, the θ wave energy feature, the α wave energy feature and the β wave energy feature are all commonly used EEG features for evaluating neural activity. When a person is tired, the above features will change. Therefore, when the prior art images neural activity for fatigue detection, the θ wave energy feature, the α wave energy feature and the β wave energy feature are usually used, or the features obtained by adding, subtracting, multiplying and dividing the θ wave energy feature, the α wave energy feature and the β wave energy feature. The above θ wave energy feature, the α wave energy feature and the β wave energy feature in the prior art usually refer to the energy features directly calculated after Fourier transforming the corresponding original brain wave spectrum (such as digital discrete Fourier transform), and more specifically refer to the corresponding power spectrum obtained after Fourier transforming the original brain wave spectrum. The energy feature can be calculated by the area enclosed by the power spectrum and the coordinate axis. The θ wave energy feature, the α wave energy feature and the β wave energy feature calculated by the prior art in the above manner are still not sensitive enough to fatigue, so the accuracy of fatigue detection is still not high enough. Fatigue detection for pilots requires a higher accuracy rate, especially during the flight of the pilot. A higher fatigue detection accuracy rate helps to detect the pilot's state more accurately, thereby facilitating timely intervention when pilot fatigue is found, thereby improving flight safety. Therefore, the inventor hopes to further improve the accuracy of fatigue detection by imaging neural activity.

[0086] First, refer to Figure 1 , an embodiment of the present invention discloses a fatigue detection method, comprising the following steps:

[0087] S1: Detect the EEG signals corresponding to multiple signal collection positions on the head.

[0088] S2: Determine the two-dimensional positions of the multiple signal collection positions according to the relative positional relationship of the multiple signal collection positions.

[0089] Exemplarily, the two-dimensional position may be in the form of a matrix, or may be in the form of a table, which is helpful for the subsequent acquisition of a digital image (electroencephalogram information image) corresponding to the electroencephalogram signal.

[0090] S3: Determine the corresponding θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics according to the EEG signals corresponding to the multiple signal collection positions on the two-dimensional position, and determine the pixel value corresponding to each signal collection position on the two-dimensional position.

[0091] Among them, the pixel value reflects the size of the θ wave energy feature, the δ wave energy feature and the α wave periodic energy feature of the EEG signal corresponding to each signal collection position in the two-dimensional position.

[0092] Specifically, digital images can be generally divided into black and white images, grayscale images, and color images according to the difference in grayscale levels. When converting EEG signals into digital images, pixel values ​​can be defined by EEG features. If a single EEG feature is selected, the obtained digital image is a monochrome image. The color change in the monochrome image is not obvious enough, which is not conducive to accurately identifying whether the image reflects fatigue or wakefulness. In this embodiment, multiple EEG features are selected to define pixel values, that is, the pixel value corresponding to the signal acquisition position is determined by the θ wave energy feature, δ wave energy feature, and α wave periodic energy feature corresponding to the EEG signal, so that a color digital image can be obtained. The color of the digital image is more vivid, and the color difference between different areas on the digital image is also more obvious, which is more conducive to improving the accuracy of fatigue detection by recognizing images.

[0093] Since the pixel value is defined by the θ wave energy feature, δ wave energy feature and α wave periodic energy feature corresponding to the EEG signal, the pixel value reflects the magnitude of the θ wave energy feature, δ wave energy feature and α wave periodic energy feature of the EEG signal corresponding to each signal acquisition position at the two-dimensional position. Exemplarily, the pixel value of each pixel point in the digital image is jointly determined by three color components (red component, green component and blue component), and the intensity of the red component, green component and blue component in each pixel value can be defined by the θ wave energy feature, δ wave energy feature and α wave periodic energy feature. Among them, in this embodiment, the θ wave energy characteristic, the δ wave energy characteristic and the α wave periodic energy characteristic can correspond one-to-one to the red component, the green component and the blue component, for example, the θ wave energy characteristic corresponds to the red component, the δ wave energy characteristic corresponds to the green component, and the α wave periodic energy characteristic corresponds to the blue component; or the θ wave energy characteristic corresponds to the red component, the α wave periodic energy characteristic corresponds to the green component, and the δ wave energy characteristic corresponds to the blue component; or the α wave periodic energy characteristic corresponds to the red component, the δ wave energy characteristic corresponds to the green component, and the θ wave energy characteristic corresponds to the blue component; or the α wave periodic energy characteristic corresponds to the red component, the θ wave energy characteristic corresponds to the green component, and the δ wave energy characteristic corresponds to the blue component; or the δ wave energy characteristic corresponds to the red component, the α wave periodic energy characteristic corresponds to the green component, and the θ wave energy characteristic corresponds to the blue component; or the δ wave energy characteristic corresponds to the red component, the α wave periodic energy characteristic corresponds to the green component, and the θ wave energy characteristic corresponds to the blue component; or the δ wave energy characteristic corresponds to the red component, the θ wave energy characteristic corresponds to the green component, and the α wave periodic energy characteristic corresponds to the blue component. In other embodiments, the θ wave energy feature, the δ wave energy feature and the α wave periodic energy feature may correspond to the red component, the green component and the blue component after operations such as addition, subtraction, multiplication and division.

[0094] S4: Determine an EEG information image according to the pixel value corresponding to each signal collection position in the two-dimensional position. The EEG information image is used for fatigue judgment.

[0095] Among them, when fatigue is judged by EEG information images, it can be judged by a person observing the images, or it can be judged by recognizing the images through a fatigue detection model.

[0096] By adopting the above technical scheme, the fatigue detection method of the present invention determines the pixel value corresponding to each signal collection position in a two-dimensional position by specifically selecting the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal, and then obtains the EEG information image used for fatigue judgment. Compared with other EEG characteristics, the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics can better reflect the user's fatigue condition. The θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics are more sensitive to fatigue, and therefore can improve the accuracy of fatigue judgment.

[0097] In the above embodiment, the alpha wave periodic energy characteristics of the EEG signal are determined by:

[0098] S01: Determine the logarithmic power spectrum of the EEG signal.

[0099] refer to Fig. 9 , Fig. 9 Curve 2 in FIG. 2 is the logarithmic power spectrum of the EEG signal.

[0100] S02: Perform nonlinear fitting based on the logarithmic power spectrum using the following formula 1 to obtain the corresponding fitting curve:

[0101] L(f)=ab×log(P(f)) (Formula 1)

[0102] Wherein, L(f) is the fitting curve, P(f) is the logarithmic power spectrum, a is the offset parameter of the fitting curve, b is the scaling transformation parameter of the fitting curve, a and b are determined by least squares fitting, f is the frequency of the EEG signal, and the range of f is 2-5 Hz and 14-30 Hz.

[0103] refer to Fig. 9 , in this embodiment, Fig. 9 Curve 1 in the figure is the fitting curve, a represents the left and right offset of curve 1, and b represents the expansion and contraction transformation of curve 1. By selecting the frequency f range of 2-5Hz and 14-30Hz during fitting, the α band is avoided, and the influence on the α band can be reduced during fitting, so as to obtain more accurate α wave band energy later.

[0104] S03: Determine the α wave frequency band energy according to the logarithmic power spectrum and the fitting curve, and the α wave periodic energy characteristic is the α wave frequency band energy.

[0105] Compared with the α wave energy characteristics used in the prior art, the α wave frequency band energy calculated by the above method is more highly correlated with fatigue, is more sensitive to fatigue, and can better reflect whether the user is tired. Therefore, the EEG information image obtained by using the α wave periodic energy characteristics is conducive to more accurate fatigue detection.

[0106] In the above embodiment, step S03 specifically includes the following steps:

[0107] S031: Determine the alpha wave spectrum periodic component according to the logarithmic power spectrum and the fitting curve using the following formula 2:

[0108] P′(f)=P(f)-L(f) (Formula 2)

[0109] Among them, P′(f) is the periodic component of the α wave spectrum, L(f) is the fitting curve, P(f) is the logarithmic power spectrum, and f is the frequency of the EEG signal.

[0110] S032: Determine the α wave spectrum periodic energy component according to the α wave spectrum periodic component by the following formula 3:

[0111] P″(f)=10×10 P′(f) (Formula 3)

[0112] Where P′(f) is the periodic component of the α wave spectrum, P″(f) is the periodic energy component of the α wave spectrum, and f is the frequency of the EEG signal. Fig.10 , Fig.10 Curve 3 in the figure is the periodic energy component of the α wave spectrum.

[0113] S033: Determine the α wave frequency band energy according to the α wave spectrum periodic energy component.

[0114] The above formula 3 can be used to obtain the periodic energy component P″(f) of the α wave spectrum, and then the energy corresponding to the α band, that is, the α wave band energy, can be calculated based on P″(f).

[0115] The present invention improves the EEG features used in constructing images and calculates the α-wave periodic energy features, δ-wave energy features and θ-wave energy features that are more sensitive to fatigue, thereby constructing an EEG information image that is more highly correlated with fatigue and improving the accuracy of fatigue judgment through images.

[0116] The energy characteristics of the θ wave and the energy characteristics of the δ wave can be obtained in the following way: perform Fourier transform (such as digital discrete Fourier transform) on the original spectrum of the EEG signal to obtain the corresponding power spectrum, and calculate the energy of the θ band and the δ band according to the area enclosed by the power spectrum and the coordinate axis to obtain the energy characteristics of the θ wave and the δ wave.

[0117] Furthermore, the inventors found through experiments that theta wave energy characteristics, delta wave energy characteristics and alpha wave periodic energy characteristics are more sensitive to fatigue. Specifically, the inventors used the following experimental methods to determine the sensitivity of different EEG characteristics to fatigue:

[0118] Select multiple signal collection positions on the head and collect awake EEG signals and fatigue EEG signals corresponding to each signal collection position, wherein the awake EEG signal is the EEG signal in the awake state, and the fatigue EEG signal is the EEG signal in the fatigue state. Subsequently, determine the multiple awake EEG features corresponding to the awake EEG signal and the multiple fatigue EEG features corresponding to the fatigue EEG signal based on the awake EEG signals and fatigue EEG signals collected at each signal collection position. Determine the degree of difference between the same type of awake EEG features and fatigue EEG features at each signal collection position, and sort the signal collection positions and the corresponding EEG features according to the degree of difference to obtain a sorting result.

[0119] Among them, the difference between the same type of awake EEG features and fatigue EEG features is determined by the feature selection algorithm, and the following five feature selection algorithms are used: minimum redundancy maximum relevance (MRMR) algorithm, chi-square test algorithm, ReliefF algorithm, Kruskal-Wallis test algorithm, and recursive feature ranking algorithm based on support vector machine (SVM-REF). The five feature selection algorithms can obtain five ranking results in total, which are recorded as rank1, rank2, rank3, rank4 and rank5 respectively.

[0120] In this experiment, a total of 61 signal acquisition positions were set according to the 10-20 international EEG lead definition, namely: 'Fp1', 'AF3', 'AF7', 'Fz', 'Fl', 'F3', 'F5', 'F7', 'FC1', 'FC3', 'FC5', 'FT7', 'Cz', 'C1', 'C3', 'C5', 'T7', 'CP1', 'CP3', 'CP5', 'TP7', 'TP9', 'Pz', 'P1', 'P3', 'P5', 'P7', 'P9' There are 9 EEG features, namely, delta wave energy feature, theta wave energy feature, alpha wave energy feature, alpha wave peak feature, beta wave energy feature, non-periodic shift feature, non-periodic stretching and transformation feature, alpha wave frequency band energy and alpha wave frequency band peak feature.

[0121] Among them, the δ wave energy feature, the θ wave energy feature, the α wave energy feature, the α wave peak feature and the β wave energy feature can all be calculated by the power spectrum obtained by digital discrete Fourier transform of the EEG signal. The offset parameter a in formula 1 is the non-periodic offset feature, and the telescopic transformation parameter b in formula 1 is the non-periodic telescopic transformation feature. The above formula 3 can be used to obtain the periodic energy component P″(f) of the α wave spectrum. The periodic energy component P″(f) of the α wave spectrum is dimensionless. Then, the energy corresponding to the α band can be calculated according to P″(f), that is, the α wave band energy, and the α wave band peak feature can also be obtained. Among them, the α wave band energy can correspond to the area enclosed by the α wave spectrum periodic energy component P″(f) and the coordinate axis in the α band (i.e., 8-13Hz), and the α wave band peak feature can correspond to the peak value of the α wave spectrum periodic energy component P″(f).

[0122] In rank1, rank2, rank3, rank4 and rank5, the number of occurrences of the above 9 features in the top 50 of each sorting result is counted. The more times each feature appears, the more important the feature is and the more sensitive it is to fatigue. For example, if the delta wave energy feature appears 10 times and the theta wave energy feature appears 6 times in the top 50 of rank1, then the importance of the delta wave energy feature is greater than that of the theta wave energy feature. By analogy, the respective advantageous features of rank1, rank2, rank3, rank4 and rank5 can be realized respectively. Combining the five sorting results, we can get the following: Fig.16 Statistical diagram of . It can be seen that the δ wave energy feature, θ wave energy feature, non-periodic stretching transformation feature, α wave frequency band energy and α wave frequency band peak feature are all suitable for fatigue detection. Among them, there is information redundancy between some types of EEG features. Considering the rapidity and simplicity of calculation, the stability, reliability and versatility of features, and the need to minimize the information redundancy between the selected features, δ wave energy features, θ wave energy features and α wave frequency band energy were finally selected to construct the EEG information image. There is little information redundancy between the δ wave energy feature, the θ wave energy feature and the α wave frequency band energy, and there is basically no information redundancy. The characteristics of EEG signals can be shown from three different angles, and they are all highly sensitive to fatigue, which helps to improve the accuracy of fatigue detection.

[0123] In the above embodiments, reference Figure 2 , step S2 specifically includes the following steps:

[0124] S21: determining an initial two-dimensional position according to the relative position relationship of multiple signal collection positions, wherein the initial two-dimensional position includes a signal collection position and a blank position, and the blank position is a position where no EEG signal is detected.

[0125] The multiple signal collection positions are determined by the positions of the electrodes placed in the 10-20 system electrode placement method. Figure 8 , Figure 8 The positions of the electrodes placed in the 10-20 system electrode placement method in this embodiment are shown. The signal acquisition position is the position where the acquisition electrode is set to detect the EEG signal. The blank position is the position where the acquisition electrode is not set to detect the EEG signal.

[0126] Furthermore, when collecting EEG signals, the more electrodes used, the better, for example, more than 16 leads (that is, the number of electrodes used), the collected EEG signals are more comprehensive, which helps to improve the accuracy of subsequent fatigue detection. In addition, when collecting EEG signals, a reference electrode (REF electrode) can be set between electrodes Fz and Cz, and a ground electrode (GND electrode) can be placed between electrodes Fpz and Fz.

[0127] Table 1 Initial two-dimensional positions of electrodes placed according to the 10-20 system electrode placement method after original expansion

[0128]

[0129] Taking the 10-20 system electrode placement method as an example, refer to Table 1, which shows the initial two-dimensional positions of the electrodes placed in the 10-20 system electrode placement method after the original expansion (i.e., initial plane expansion), for example, the grid in the first row and first column corresponds to the blank position, and the grid in the first row and fourth column corresponds to the signal collection position. When the electrodes are set on the head, the surface formed by the multiple electrodes is in the form of a three-dimensional spherical surface. After the plane expansion, some areas are enlarged, and some areas are not covered by the electrodes, so blank positions will appear.

[0130] S22: In the initial two-dimensional position, fill the blank position on the left side of the middle column of the initial two-dimensional position with the type of the position on the right side of the blank position, fill the blank position on the right side of the middle column of the initial two-dimensional position with the type of the position on the left side of the blank position, and fill the blank position in the middle column of the initial two-dimensional position with the type of the position relatively closer to the center of the middle column among the positions above and below the blank position, to obtain the two-dimensional position.

[0131] Exemplarily, the blank positions are filled in the initial two-dimensional positions shown in Table 1 to obtain the two-dimensional positions shown in Table 2. Table 2 shows the two-dimensional positions after the positions of the electrodes placed in the 10-20 system electrode placement method are corrected and expanded (i.e., the two-dimensional positions after the blank positions are filled). For example, for the blank positions corresponding to the third column of the first row in Table 1, the type Fp1 of the signal acquisition position located on its right side, i.e., the fourth column of the first row, is filled in; for the blank positions corresponding to the eighth column of the second row in Table 1, the type AF8 of the signal acquisition position located on its left side, i.e., the seventh column of the second row, is filled in; for the blank positions corresponding to the fifth column of the fourth row, the type of position relatively closer to the center of the middle column is filled in. The signal acquisition position Cz of the fifth column of the fifth row is closer to the center of the middle column than the signal acquisition position Fz of the fifth column of the third row, so Cz is filled in the blank positions corresponding to the fifth column of the fourth row. In addition, for the blank positions of the first column of the ninth row in Table 1, the blank positions located on its right side (i.e., the blank positions of the second column of the ninth row) should be filled in first, and the same is true for the rest of the blank positions similar to the first column of the ninth row in Table 1.

[0132] Table 2 Two-dimensional position after correction and expansion according to the position of the electrodes placed in the 10-20 system electrode placement method

[0133]

[0134] In the process of converting neural activity into digital images, if the blank positions in the initial two-dimensional positions are not filled, and the pixel values ​​corresponding to the blank positions are directly set to 0, there will be local black pixels in the final EEG information image, which is not convenient for observation and is not conducive to subsequent fatigue detection through EEG information images. The above method is used to fill the blank positions in the initial two-dimensional positions, which can avoid the local pixels in the generated EEG information image being black, reduce singular values, and improve the stability of the data. In addition, the brain has left-right symmetry, and the signal in the middle brain area is more stable. By filling the blank positions with the middle column and the center of the middle column as the reference, the final two-dimensional position is more in line with the physiological characteristics of the EEG signal, and the final EEG information image is also smoother, which is conducive to improving the accuracy of image recognition and fatigue detection.

[0135] In the above embodiments, reference Figure 3 , the pixel value corresponding to each signal acquisition position in the two-dimensional position is determined by the following method:

[0136] S301: Determine the corresponding red channel EEG energy features, green channel EEG energy features and blue channel EEG energy features according to the θ wave energy features, δ wave energy features and α wave periodic energy features corresponding to each signal collection position on the two-dimensional position.

[0137] Furthermore, step S301 specifically includes the following steps:

[0138] S3011: respectively determine that the red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are one of the following three types: theta wave energy feature, delta wave energy feature, alpha wave periodic energy feature, and the red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are different from each other.

[0139] Specifically, there are six one-to-one correspondences between the red channel EEG energy characteristics, the green channel EEG energy characteristics, the blue channel EEG energy and the θ wave energy characteristics, the δ wave energy characteristics, and the α wave periodic energy characteristics. For example, one of them may be to determine that the red channel EEG energy characteristics are the θ wave energy characteristics, the green channel EEG energy characteristics are the δ wave energy characteristics, and the blue channel EEG energy characteristics are the α wave periodic energy characteristics. The remaining five can be obtained similarly and are not listed here one by one.

[0140] The above method is used to determine the red channel EEG energy characteristics, green channel EEG energy characteristics and blue channel EEG energy characteristics. The calculation method is simple, and the finally generated EEG information image has bright colors and high visual distinction of the human eye. It is not only suitable for accurately detecting fatigue through fatigue detection models, but also relevant personnel can accurately judge whether the person being tested is fatigued by observing the EEG information image.

[0141] S302: Normalizing the red channel EEG energy feature, the green channel EEG energy feature, and the blue channel EEG energy feature respectively to obtain a red channel pixel value, a green channel pixel value, and a blue channel pixel value.

[0142] In this embodiment, in each pixel value (r, g, b), the value range of the red channel pixel value r, the green channel pixel value g and the blue channel pixel value b is between 0-1, while the difference between the θ wave energy feature, the δ wave energy feature and the α wave periodic energy feature is large, and the range of variation is wide. Usually, the maximum values ​​of the θ wave energy feature and the δ wave energy feature are about 30, while the maximum value of the α wave periodic energy feature is about 100, which does not meet the input requirements. There is a risk of exceeding the pixel value range when directly constructing an image based on the red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature. Therefore, normalization is required to obtain red channel pixel values, green channel pixel values ​​and blue channel pixel values ​​between 0-1. In addition, the calculation method of the α wave periodic energy feature is different from that of the θ wave energy feature and the δ wave energy feature, and there are also differences in units. Therefore, it is also necessary to normalize the three EEG features through a unified standard to eliminate the influence of unit differences.

[0143] S303: Determine a pixel value according to the red channel pixel value, the green channel pixel value and the blue channel pixel value.

[0144] The red channel pixel value, green channel pixel value and blue channel pixel value obtained after normalization are all within the range of 0-1, so the pixel value can be determined according to the red channel pixel value, green channel pixel value and blue channel pixel value, and then the EEG information image can be obtained.

[0145] In the above embodiment, step S302 may specifically include the following steps:

[0146] S3021: Normalize the red channel EEG energy feature, the green channel EEG energy feature, and the blue channel EEG energy feature respectively by the following formula 4, formula 5, and formula 6 to obtain the red channel pixel value, the green channel pixel value, and the blue channel pixel value:

[0147] r′ i =((r i -min(r)) / (max(r)-min(r)) (Formula 4)

[0148] g′ i =((g i -min(g)) / (max(g)-min(g)) (Formula 5)

[0149] b′ i =((b i-min(b)) / (max(b)-min(b)) (Formula 6)

[0150] Among them, r i represents the red channel EEG energy feature corresponding to the EEG signal measured at the i-th signal acquisition position, r′ i Indicates r i Normalized red channel EEG energy feature, g i represents the green channel EEG energy feature corresponding to the EEG signal measured at the i-th signal acquisition position, g′ i Indicates g i Normalized green channel EEG energy characteristics, b i represents the blue channel EEG energy feature corresponding to the EEG signal measured at the i-th signal acquisition position, b′ i Indicates b i The normalized blue channel EEG energy feature, max(r) represents the maximum value in the set of multiple known red channel EEG energy features, min(r) represents the minimum value in the set of multiple known red channel EEG energy features, max(g) represents the maximum value in the set of multiple known green channel EEG energy features, min(g) represents the minimum value in the set of multiple known green channel EEG energy features, max(b) represents the maximum value in the set of multiple known blue channel EEG energy features, and min(b) represents the minimum value in the set of multiple known blue channel EEG energy features.

[0151] Among them, a set of multiple known red channel EEG energy features, a set of multiple known green channel EEG energy features, and a set of multiple known blue channel EEG energy features can be determined by:

[0152] The EEG signal corresponding to each signal acquisition position has a certain time length, and the EEG signal can be segmented, and the step segmentation is performed according to a certain step length, and each segment of the separated EEG signal has the same time length. Exemplarily, the time length of each segment of the EEG signal is a preset time length, and the preset time length can be, for example, 2-10 seconds, and the step length can be selected to be 10%-50% of the time length of the EEG signal. For example, the length of each segment of the separated EEG signal can be 4 seconds, and the step length can be 2 seconds, that is, the first second to the fifth second is the first EEG signal segment, the third second to the seventh second is the second EEG signal segment, and so on. Subsequently, the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics corresponding to each EEG signal segment can be obtained, and then the corresponding red channel EEG energy characteristics, green channel EEG energy characteristics and blue channel EEG energy characteristics can be determined. The set of multiple red channel EEG energy features is the set of the multiple red channel EEG energy features known above, the set of multiple green channel EEG energy features is the set of the multiple green channel EEG energy features known above, and the set of multiple blue channel EEG energy features is the set of the multiple blue channel EEG energy features known above.

[0153] max(r), min(r), max(g), min(g), max(b) and min(b) are all determined in advance from the above-mentioned known sets. Further, the set of known multiple red channel EEG energy features, the set of known multiple green channel EEG energy features and the set of known multiple blue channel EEG energy features can all be sets constructed based on the EEG signals of the same person (for example, the person who is currently actually detecting the EEG signal). In other embodiments, the set of known multiple red channel EEG energy features, the set of known multiple green channel EEG energy features and the set of known multiple blue channel EEG energy features can also all be sets constructed based on the EEG signals of multiple people, which are universal and suitable for most people. Preferably, when the set of known multiple red channel EEG energy features, the set of known multiple green channel EEG energy features and the set of known multiple blue channel EEG energy features are all sets constructed based on the EEG signals of the same person (for example, the person who is currently actually detecting the EEG signal), normalization is more accurate and more targeted, which helps to further improve the accuracy of fatigue detection.

[0154] By determining the maximum and minimum values ​​in a set of known multiple red channel EEG energy features, the maximum and minimum values ​​in a set of known multiple green channel EEG energy features, and the maximum and minimum values ​​in a set of known multiple blue channel EEG energy features, it helps to perform more accurate normalization later and helps to improve the accuracy of fatigue detection through the generated images.

[0155] Further, the known set of multiple red channel EEG energy features includes multiple red channel EEG energy features corresponding to awake EEG signals and multiple red channel EEG energy features corresponding to fatigue EEG signals, the known set of multiple green channel EEG energy features includes multiple green channel EEG energy features corresponding to awake EEG signals and multiple green channel EEG energy features corresponding to fatigue EEG signals, and the known set of multiple blue channel EEG energy features includes multiple blue channel EEG energy features corresponding to awake EEG signals and multiple blue channel EEG energy features corresponding to fatigue EEG signals. In this way, the EEG features corresponding to the awake state and the EEG features corresponding to the fatigue state are considered together, and the maximum and minimum values ​​are selected. When normalized by a unified standard, the difference of the same type of EEG features corresponding to the awake state and the fatigue state can be retained, so that the generated EEG information image corresponding to the awake state and the EEG information image corresponding to the fatigue state can have a relatively obvious difference, and can also reduce the order of magnitude, and normalize the pixel values ​​of different color channels to between 0 and 1, which helps to improve the accuracy of image recognition, and then improve the accuracy of fatigue detection.

[0156] Furthermore, in other embodiments, the set of known multiple red channel EEG energy features can also be composed of red channel EEG energy features corresponding to multiple awake EEG signals and red channel EEG energy features corresponding to multiple fatigue EEG signals, the set of known multiple green channel EEG energy features can be composed of green channel EEG energy features corresponding to multiple awake EEG signals and green channel EEG energy features corresponding to multiple fatigue EEG signals, and the set of known multiple blue channel EEG energy features can be composed of blue channel EEG energy features corresponding to multiple awake EEG signals and blue channel EEG energy features corresponding to multiple fatigue EEG signals.

[0157] In the above embodiment, in order to further enhance the color difference between different regions on the same image and the difference between different images, step S302 specifically further includes the following steps:

[0158] S3022: Based on the normalized results obtained by Formula 4, Formula 5 and Formula 6, normalize again by the following Formula 7, Formula 8 and Formula 9 to obtain the red channel pixel value, the green channel pixel value and the blue channel pixel value:

[0159]

[0160] Among them, r″ i represents the final normalized red channel EEG energy feature corresponding to the i-th signal acquisition position, g″ i represents the final normalized green channel EEG energy feature corresponding to the i-th signal acquisition position, b″i represents the final normalized blue channel EEG energy feature corresponding to the i-th signal acquisition position, r′ i represents the red channel EEG energy feature normalized by Formula 4 corresponding to the i-th signal acquisition position, g′ i represents the green channel EEG energy feature normalized by Formula 5 corresponding to the i-th signal acquisition position, b′ i represents the blue channel EEG energy feature normalized by Formula 6 corresponding to the i-th signal acquisition position, and k is 7-13. Preferably, k is 10, at which time the normalized EEG information image has a better visual effect and is more conducive to image recognition, thereby improving the accuracy of fatigue detection.

[0161] In practical applications, the red channel EEG energy features at different signal acquisition positions are not necessarily between (max(r)-min(r)), and the green channel EEG energy features and the blue channel EEG energy features are similar. Therefore, if normalization is performed only through Formula 4, Formula 5, and Formula 6, there may still be a risk that the pixel value is not in the range of 0-1. By adopting Formula 7, Formula 8, and Formula 9, the red channel pixel value, the green channel pixel value, and the blue channel pixel value can be more strictly controlled between 0-1, further improving the stability of the data, while improving the differences between different regions on the same image and the differences between different images, further improving the convenience of image recognition, and improving the accuracy and reliability of fatigue detection.

[0162] In the above-mentioned normalization step, the red channel EEG energy features, the green channel EEG energy features and the blue channel EEG energy features are normalized respectively by a unified standard (it can be normalized only by step S3021, or it can be normalized by both steps S3021 and S3022), which can narrow the differences between the EEG energy features of different color channels, prevent the subsequent pixel value of a certain color channel from being too large, causing the pixel values ​​of the other color channels to be basically ignored, and make the pixel values ​​of the three color channels have non-negligible contributions to the pixel value of each pixel point, so that the generated EEG information image is more balanced and the color distinction is more obvious, which is helpful for subsequent image recognition and improves the accuracy of fatigue detection through images.

[0163] A more specific embodiment is introduced below by taking the red channel EEG energy feature corresponding to the δ wave energy feature, the green channel EEG energy feature corresponding to the α wave periodic energy feature, and the blue channel EEG energy feature corresponding to the θ wave energy feature as an example. For ease of description, the red channel EEG energy feature r is replaced by the letter δ, the green channel EEG energy feature g is replaced by the letter α, and the blue channel EEG energy feature b is replaced by the letter θ. Before normalization, the image features initially constructed based on Table 2 (i.e., the pixel values ​​corresponding to each position in the two-dimensional position) are shown in Table 3 below.

[0164] Table 3 Image features constructed based on Table 2

[0165]

[0166] Normalize according to the following formulas 4-1, 5-1 and 6-1:

[0167] δ′ i =((δ i -min(δ)) / (max(δ)-min(δ)) (Formula 4-1)

[0168] α′ i =((α i -min(α)) / (max(α)-min(α)) (Formula 5-1)

[0169] θ′ i =((θ i -min(θ)) / (max(θ)-min(θ)) (Formula 6-1)

[0170] Among them, δ′ i represents the normalized delta wave energy characteristic δ corresponding to the i-th signal acquisition position i , α′ i represents the normalized α wave periodic energy characteristic α corresponding to the i-th signal acquisition position i ,θ′ i represents the normalized θ wave energy feature θ corresponding to the i-th signal acquisition position i max(δ) represents the maximum value of the δ wave energy feature of the EEG signal, and min(δ) represents the minimum value of the δ wave energy feature of the EEG signal. max(α) represents the maximum value of the α wave periodic energy feature of the EEG signal, and mmin(α) represents the minimum value of the α wave periodic energy feature of the EEG signal. max(θ) represents the maximum value of the θ wave energy feature of the EEG signal, and mmin(θ) represents the minimum value of the θ wave energy feature of the EEG signal.

[0171] Among them, δ i represents the delta wave energy characteristic corresponding to the EEG signal measured at the i-th signal acquisition position, δ′ i Indicates the i Normalized delta wave energy characteristics, α i represents the periodic energy characteristics of the α wave corresponding to the EEG signal measured at the i-th signal acquisition position, α′ i Indicates α i Normalized α wave periodic energy characteristics, θ i represents the θ wave energy characteristic corresponding to the EEG signal measured at the i-th signal acquisition position, θ′i Represents the i The normalized θ wave energy characteristics, max(δ) represents the maximum value in the set of multiple known δ wave energy characteristics, mmin(δ) represents the minimum value in the set of multiple known δ wave energy characteristics, max(α) represents the maximum value in the set of multiple known α wave periodic energy characteristics, mmin(α) represents the minimum value in the set of multiple known α wave periodic energy characteristics, max(θ) represents the maximum value in the set of multiple known θ wave energy characteristics, and rmin(θ) represents the minimum value in the set of multiple known θ wave energy characteristics.

[0172] Among them, the known set of multiple delta wave energy features, the known set of multiple alpha wave periodic energy features, and the known set of multiple theta wave energy features are obtained in the same manner as the known set of multiple red channel EEG energy features, the known set of multiple green channel EEG energy features, and the known set of multiple blue channel EEG energy features in the aforementioned embodiment. Further, the known set of multiple delta wave energy features includes multiple delta wave energy features corresponding to wakefulness and multiple delta wave energy features corresponding to fatigue, the known set of multiple alpha wave periodic energy features includes multiple alpha wave periodic energy features corresponding to wakefulness and multiple alpha wave periodic energy features corresponding to fatigue, and the known set of multiple theta wave energy features includes multiple theta wave energy features corresponding to wakefulness and multiple theta wave energy features corresponding to fatigue.

[0173] Further, the set of known multiple delta wave energy features, the set of known multiple alpha wave periodic energy features, and the set of known multiple theta wave energy features may all be sets constructed based on the EEG signals of the same person (e.g., the person currently actually detecting the EEG signals). In other embodiments, the set of known multiple delta wave energy features, the set of known multiple alpha wave periodic energy features, and the set of known multiple theta wave energy features may all be sets constructed based on the EEG signals of multiple people, which are universal and suitable for most people. Preferably, when the set of known multiple delta wave energy features, the set of known multiple alpha wave periodic energy features, and the set of known multiple theta wave energy features are all sets constructed based on the EEG signals of the same person (e.g., the person currently actually detecting the EEG signals), normalization is more accurate and more targeted, which helps to further improve the accuracy of fatigue detection.

[0174] After the above normalization steps, we can finally get Fig.11 A set of awake and fatigue EEG information images (including nine tests) is shown. After sufficient training of the fatigue detection model, the accuracy of the recognition of EEG information images can be improved. Fig.11In the EEG information image shown, the red channel EEG energy feature corresponds to the alpha wave periodic energy feature, the green channel EEG energy feature corresponds to the theta wave energy feature, and the blue channel EEG energy feature corresponds to the delta wave energy feature. In this case, the difference between some awake and tired images is still not obvious enough when observed with the naked eye, and further normalization can be performed on this basis.

[0175] Then, normalization is performed according to the following formulas 7-1, 8-1 and 9-1:

[0176]

[0177] Among them, δ″ i represents the final normalized red channel EEG energy feature corresponding to the i-th signal acquisition position, α″ i represents the final normalized green channel EEG energy feature corresponding to the i-th signal acquisition position, θ″ i represents the final normalized blue channel EEG energy feature corresponding to the i-th signal acquisition position, δ′ i represents the red channel EEG energy feature normalized by formula 4-1 corresponding to the i-th signal acquisition position, α′ i represents the green channel EEG energy feature normalized by formula 5-1 corresponding to the i-th signal acquisition position, θ′ i Represents the blue channel EEG energy characteristics normalized by formula 6-1 corresponding to the i-th signal acquisition position.

[0178] After the above two normalizations, the following table 4 can be obtained, and then the EEG information image can be obtained.

[0179] Table 4 Pixel values ​​of each signal acquisition position after normalization of Table 3

[0180]

[0181] For example, when the red channel EEG energy feature corresponds to the delta wave energy feature, the green channel EEG energy feature corresponds to the alpha wave periodic energy feature, and the blue channel EEG energy feature corresponds to the theta wave energy feature, after the above two normalizations, the final awake and fatigue EEG information images are as follows: Fig.12 shown.

[0182] When the red channel EEG energy feature corresponds to the δ wave energy feature, the green channel EEG energy feature corresponds to the θ wave energy feature, and the blue channel EEG energy feature corresponds to the α wave periodic energy feature, after the above two normalizations, the final EEG information images of wakefulness and fatigue are as follows: Fig.13 As shown (including nine tests).

[0183] When the red channel EEG energy feature corresponds to the alpha wave periodic energy feature, the green channel EEG energy feature corresponds to the theta wave energy feature, and the blue channel EEG energy feature corresponds to the delta wave energy feature, after the above two normalizations, the final awake and fatigue EEG information images are as follows: Fig.14 As shown (including nine tests).

[0184] It can be seen that although there are differences in the colors of the EEG information images obtained in the above three situations, the differences between the EEG information images corresponding to fatigue and wakefulness are very obvious. Human naked eye observation and fatigue detection models can more accurately identify whether the image corresponds to a fatigue state or a wakefulness state. The other three situations included in the above step S3011 have similar effects and will not be repeated here.

[0185] In the above embodiments, reference Figure 5 The aforementioned step S1 may specifically include the following steps:

[0186] S11: Detecting the original EEG signals corresponding to the multiple signal collection positions on the head.

[0187] S12: Preprocess the original EEG signal to obtain an EEG signal.

[0188] Among them, preprocessing includes one or more of the following methods: filtering, removing power frequency interference, bad channel detection and interpolation, independent component analysis, re-referencing, data cleaning and screening.

[0189] More specifically, after reading the original EEG signal, the original EEG signal can be filtered to remove unnecessary frequency components, such as removing low-frequency DC components and high-frequency noise. In this embodiment, bandpass filtering can be used, and the bandpass frequency is 0.5-45Hz, so as to retain the required components of the EEG signal, which is conducive to extracting EEG features for fatigue detection. After reading the original EEG signal, the original EEG signal can also be removed from the power frequency interference, and more specifically, the power frequency interference can be removed by notch filtering to improve the accuracy of the original EEG signal; bad channel detection and interpolation can also be performed. The bad channel refers to the position of the electrode where the collected signal quality is poor. Interpolation refers to using the EEG signal of the adjacent channel of the bad channel to estimate the missing EEG signal at the bad channel, so that the missing signal can be improved and repaired.

[0190] In preprocessing, independent component analysis (ICA) can be used to remove physiological and non-physiological artifacts such as eye movements and electrocardiogram.

[0191] Re-referencing refers to converting the original EEG signal data from one reference point to another reference point to reduce noise. One reference point can be the reference electrode set when collecting EEG signals, and the other reference point can be the average value of multiple collected EEG signals.

[0192] Data cleaning and screening can include removing outliers and removing bad test data. Among them, removing outliers means cutting off the signal when it exceeds a certain threshold. Removing bad test data means identifying and excluding test data contaminated by artifacts, as well as test data that are not suitable for further analysis for various reasons.

[0193] Furthermore, the accuracy of the EEG signal can be improved by selecting one of the above methods for preprocessing, and the accuracy of the EEG signal can be further improved by using multiple methods. In the process of preprocessing the original EEG signal, when filtering, removing power frequency interference, bad channel detection and interpolation, independent component analysis, re-referencing and data cleaning and screening are all adopted, the order of filtering, removing power frequency interference and bad channel detection and interpolation is not limited and can be interchanged, and re-referencing is performed after independent component analysis.

[0194] Furthermore, when fatigue detection is performed by using a fatigue detection model to identify EEG information images, data segmentation may also be involved in the preprocessing process to increase the number of samples, which is helpful in training a fatigue detection model with better performance and improving the accuracy of fatigue detection by the fatigue detection model. Exemplarily, when data is segmented, the original EEG signal can be sliced ​​in steps according to a certain length, with the length of each data segment being 2-10 seconds, and the step length can be selected to be 10%-50% of the data length. For example, the length of each data segment can be 6 seconds, and the corresponding step can be, for example, 2 seconds.

[0195] Further, refer to Figure 4 , the fatigue detection method also includes the following steps:

[0196] S5: Perform image recognition on the EEG information image through the fatigue detection model to determine whether the EEG information image represents a wakeful state or a fatigued state.

[0197] Among them, the deep learning models used for image recognition can be used in the present invention to construct the above-mentioned fatigue detection model. When constructing and training the fatigue detection model, the transfer learning method can be preferentially used for deep learning. The specific steps include: on the basis of the existing machine learning model, the EEG information image is resized to meet the input requirements of the model. In addition, the last fully connected layer of the existing machine learning model can be replaced, and the learning parameters can be increased so that the output types of the model are two, one for wakefulness and the other for fatigue. Finally, model training and application testing can be performed based on database images.

[0198] The present invention uses a public dataset to test the effectiveness of the present invention according to the paper "a resting-state EEG dataset for sleep deprivation". The above paper was published in the journal Scientific Data in 2024, disclosing a dataset that has been recognized by the academic community through rigorous experimental design and a large number of samples. Its main contribution is that it provides awake and fatigue EEG data that can be used by researchers. The paper uses 61 Ag / AgCl active electrodes and arranges the electrodes according to the extended 10-20 international electrode placement system. When verifying the effectiveness of the present invention, the data in the dataset is processed according to the steps in the aforementioned embodiments, including: preprocessing (filtering, removing power frequency interference, bad channel detection and interpolation, independent component analysis, re-reference, data cleaning and screening), data segmentation, constructing two-dimensional positions, calculating the three EEG features used in the present invention, and normalizing the EEG features twice to obtain EEG information images. In addition, in order to facilitate display observation, the picture is also pixel enlarged and adjusted. The 9 randomly selected pictures after adjustment are as follows, for example Fig.12 From the set of images shown (corresponding to nine tests), it can be seen that there are differences in the EEG information images corresponding to fatigue and wakefulness.

[0199] More specifically, we also used 5398 awake images and 5266 fatigue images from the database in the above paper, and then used the pre-trained Xceptioncv network for transfer learning to build the above fatigue detection model. We used 70% of the selected data as the training set and the remaining 30% as the test set. The confusion matrix of the test set is as follows: Fig.15 As shown. Among them, it can be seen from the confusion matrix that the classification accuracy is 0.99281, the sensitivity is 0.99433, the specificity is 0.99131, the precision is 0.99121, the recall rate is 0.99433, and the f1Score is 0.99277. The above multiple indicators are all close to 1. It can be seen that the EEG information image obtained by the above method of the present invention can achieve more accurate fatigue detection.

[0200] Second, reference Figure 6 The embodiment of the present invention discloses a fatigue detection device 1, which performs fatigue detection by using any of the fatigue detection methods in the above embodiments. The fatigue detection device 1 includes an EEG signal detection module 11, a two-dimensional position determination module 12, a pixel value determination module 13 and an EEG information image determination module 14.

[0201] The EEG signal detection module 11 is connected to the two-dimensional position determination module 12 , the two-dimensional position determination module 12 is connected to the pixel value determination module 13 , and the pixel value determination module 13 is connected to the EEG information image determination module 14 .

[0202] Continue to refer Figure 6 The EEG signal detection module 11 is used to detect EEG signals corresponding to multiple signal collection positions on the head.

[0203] The two-dimensional position determination module 12 is used to determine the two-dimensional positions of multiple signal collection positions according to the relative position relationship of the multiple signal collection positions.

[0204] The pixel value determination module 13 is used to determine the corresponding θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics according to the EEG signals corresponding to the multiple signal acquisition positions on the two-dimensional position, and determine the pixel value corresponding to each signal acquisition position on the two-dimensional position. The pixel value reflects the magnitude of the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal corresponding to each signal acquisition position on the two-dimensional position.

[0205] The EEG information image determination module 14 is used to determine the EEG information image according to the pixel value corresponding to each signal collection position in the two-dimensional position. The EEG information image is used for fatigue judgment.

[0206] By adopting the above technical scheme, the fatigue detection device 1 of the present invention determines the pixel value corresponding to each signal collection position in a two-dimensional position by specifically selecting the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal, and then obtains the EEG information image used for fatigue judgment. The θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics can better reflect the user's fatigue condition than other EEG characteristics, and therefore can improve the accuracy of fatigue judgment.

[0207] Furthermore, the EEG signal detection module 11 can also be used to perform the aforementioned steps S11, S12 and related steps. The two-dimensional position determination module 12 can also be used to perform the aforementioned steps S21, S22 and related steps. The pixel value determination module 13 can also be used to perform the aforementioned steps S301, S302, S303 and related steps, and can also be used to determine the alpha wave periodic energy characteristics of the EEG signal, including the aforementioned steps S01, S02, S03, S031, S032, S033, S3011, S3021 and S3022.

[0208] Furthermore, the fatigue detection device 1 further comprises a fatigue detection module (not shown in the figure), and the fatigue detection module is used to execute the aforementioned step S5 and related steps.

[0209] Thirdly, reference Figure 7The embodiment of the present invention discloses an electronic device 2, including a processor 22 and a memory 21, wherein the memory 21 stores at least one instruction, and when the at least one instruction is executed by the processor 22, the fatigue detection method in any embodiment of the first aspect is implemented. The memory 21 may include, for example, a system memory, a fixed non-volatile storage medium, etc. The system memory may store, for example, an operating system, an application, a boot loader (BootLoader), and other programs.

[0210] In this embodiment, the electronic device 2 determines the pixel value corresponding to each signal collection position in a two-dimensional position by specifically selecting the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal, and then obtains an EEG information image for fatigue judgment. Compared with other EEG characteristics, the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics can better reflect the user's fatigue condition, and therefore can improve the accuracy of fatigue judgment.

[0211] In a fourth aspect, an embodiment of the present invention discloses a computer-readable storage medium, in which at least one instruction is stored. When the at least one instruction is executed, the fatigue detection method in any embodiment of the first aspect is implemented.

[0212] In this embodiment, the computer-readable storage medium determines the pixel value corresponding to each signal acquisition position in the two-dimensional position by specifically selecting the θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics of the EEG signal, and then obtains the EEG information image used for fatigue judgment. The θ wave energy characteristics, δ wave energy characteristics and α wave periodic energy characteristics can better reflect the user's fatigue status compared with other EEG characteristics, and therefore can improve the accuracy of fatigue judgment.

[0213] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0214] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0215] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0216] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0217] Although the present invention has been illustrated and described with reference to certain preferred embodiments of the present invention, it should be understood by those skilled in the art that the above is a further detailed description of the present invention in conjunction with specific embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. Those skilled in the art may make various changes in form and details, including making several simple deductions or substitutions, without departing from the spirit and scope of the present invention.

Claims

1. A fatigue detection method, characterized in that: include: Detecting EEG signals corresponding to multiple signal collection positions on the head; Determining the two-dimensional positions of the multiple signal collection positions according to the relative positional relationship of the multiple signal collection positions; Determine the corresponding θ wave energy feature, δ wave energy feature and α wave periodic energy feature according to the EEG signals corresponding to each of the plurality of signal acquisition positions on the two-dimensional position, and determine the pixel value corresponding to each of the signal acquisition positions on the two-dimensional position; wherein the pixel value reflects the magnitude of the θ wave energy feature, δ wave energy feature and α wave periodic energy feature of the EEG signal corresponding to each of the signal acquisition positions on the two-dimensional position; The electroencephalogram information image is determined according to the pixel value corresponding to each signal collection position on the two-dimensional position; the electroencephalogram information image is used for fatigue judgment.

2. The fatigue detection method according to claim 1, characterized in that: The alpha wave periodic energy feature of the EEG signal is determined by: Determining the logarithmic power spectrum of the EEG signal; According to the logarithmic power spectrum, a nonlinear fitting is performed using the following formula 1 to obtain the corresponding fitting curve: L(f)=ab×log(P(f))(Formula 1) Wherein, L(f) is the fitting curve, P(f) is the logarithmic power spectrum, a is the offset parameter of the fitting curve, b is the scaling transformation parameter of the fitting curve, a and b are determined by least squares fitting, f is the frequency of the EEG signal, and the range of f is 2-5 Hz and 14-30 Hz; The α wave frequency band energy is determined according to the logarithmic power spectrum and the fitting curve, and the α wave periodic energy characteristic is the α wave frequency band energy.

3. The fatigue detection method according to claim 2, characterized in that: Determining the alpha wave frequency band energy according to the logarithmic power spectrum and the fitting curve specifically includes: The alpha wave spectrum periodic component is determined according to the logarithmic power spectrum and the fitting curve by the following formula 2: P ′ (f) = P(f) - L(f) (Formula 2) Among them, P ′ (f) is the periodic component of the α wave spectrum, L(f) is the fitting curve, P(f) is the logarithmic power spectrum, and f is the frequency of the EEG signal; The α wave spectrum periodic energy component is determined according to the α wave spectrum periodic component by the following formula 3: P ″ (f) = 10 × 10 P′(f) (Formula 3) Among them, P ′ (f) is the periodic component of the α wave spectrum, P ″ (f) is the periodic energy component of the α wave spectrum, and f is the frequency of the EEG signal; The α wave frequency band energy is determined based on the α wave spectrum period energy component.

4. The fatigue detection method according to claim 1, characterized in that: Determining the two-dimensional positions of the plurality of signal collection positions according to the relative positional relationship of the plurality of signal collection positions specifically includes: Determining an initial two-dimensional position according to the relative position relationship of the plurality of signal collection positions; the initial two-dimensional position includes the signal collection position and a blank position, the blank position being a position where no EEG signal is detected; In the initial two-dimensional position, the blank position on the left side of the middle column of the initial two-dimensional position is filled with the type of the position on the right side of the blank position, the blank position on the right side of the middle column of the initial two-dimensional position is filled with the type of the position on the left side of the blank position, and the blank position in the middle column of the initial two-dimensional position is filled with the type of the position located above and below the blank position that is relatively closer to the center of the middle column, to obtain the two-dimensional position.

5. The fatigue detection method according to claim 1, characterized in that: The multiple signal collection positions are determined by the positions of electrodes placed in the 10-20 system electrode placement method.

6. The fatigue detection method according to claim 1, characterized in that: The pixel value corresponding to each signal acquisition position on the two-dimensional position is determined by the following method: Determine the corresponding red channel EEG energy feature, green channel EEG energy feature and blue channel EEG energy feature according to the θ wave energy feature, the δ wave energy feature and the α wave periodic energy feature corresponding to each signal collection position on the two-dimensional position; Normalizing the red channel EEG energy feature, the green channel EEG energy feature, and the blue channel EEG energy feature respectively to obtain a red channel pixel value, a green channel pixel value, and a blue channel pixel value; The pixel value is determined according to the red channel pixel value, the green channel pixel value and the blue channel pixel value.

7. The fatigue detection method according to claim 6, characterized in that: The step of normalizing the red channel EEG energy feature, the green channel EEG energy feature, and the blue channel EEG energy feature to obtain a red channel pixel value, a green channel pixel value, and a blue channel pixel value specifically includes: The red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are respectively normalized by the following formula 4, formula 5 and formula 6 to obtain the red channel pixel value, the green channel pixel value and the blue channel pixel value: r′ i =((r i -min(r)) / (max(r)-min(r)) (Formula 4) g′ i =((g i -min(g)) / (max(g)-min(g)) (Formula 5) b′ i =((b i -min(b)) / (max(b)-min(b)) (Formula 6) Among them, r i represents the red channel EEG energy feature corresponding to the EEG signal measured at the i-th signal acquisition position, r′ i Indicates r i Normalized red channel EEG energy feature, g i represents the green channel EEG energy feature corresponding to the EEG signal measured at the i-th signal acquisition position, g′ i Indicates g i Normalized green channel EEG energy characteristics, b i represents the blue channel EEG energy feature corresponding to the EEG signal measured at the i-th signal acquisition position, b′ i Indicates b i The normalized blue channel EEG energy feature, max(r) represents the maximum value in the set of multiple known red channel EEG energy features, min(r) represents the minimum value in the set of multiple known red channel EEG energy features, max(g) represents the maximum value in the set of multiple known green channel EEG energy features, min(g) represents the minimum value in the set of multiple known green channel EEG energy features, max(b) represents the maximum value in the set of multiple known blue channel EEG energy features, and min(b) represents the minimum value in the set of multiple known blue channel EEG energy features.

8. The fatigue detection method according to claim 7, characterized in that: The set of multiple known red channel EEG energy features includes red channel EEG energy features corresponding to multiple awake EEG signals and red channel EEG energy features corresponding to multiple fatigue EEG signals. The set of multiple known green channel EEG energy features includes green channel EEG energy features corresponding to multiple awake EEG signals and green channel EEG energy features corresponding to multiple fatigue EEG signals. The set of multiple known blue channel EEG energy features includes blue channel EEG energy features corresponding to multiple awake EEG signals and blue channel EEG energy features corresponding to multiple fatigue EEG signals.

9. The fatigue detection method according to claim 7, characterized in that: The step of normalizing the red channel EEG energy feature, the green channel EEG energy feature, and the blue channel EEG energy feature to obtain a red channel pixel value, a green channel pixel value, and a blue channel pixel value also includes: Based on the normalized results obtained by the formula 4, the formula 5 and the formula 6, normalization is performed again by the following formula 7, formula 8 and formula 9 to obtain the red channel pixel value, the green channel pixel value and the blue channel pixel value: Among them, r″ i represents the final normalized red channel EEG energy feature corresponding to the i-th signal acquisition position, g″ i represents the final normalized green channel EEG energy feature corresponding to the i-th signal acquisition position, b″ i represents the final normalized blue channel EEG energy feature corresponding to the i-th signal acquisition position, r′ i represents the red channel EEG energy feature normalized by Formula 4 corresponding to the i-th signal acquisition position, g′ i represents the green channel EEG energy feature normalized by Formula 5 corresponding to the i-th signal acquisition position, b′ i represents the blue channel EEG energy feature normalized by the formula 6 corresponding to the ith signal acquisition position, and k is 7-13.

10. The fatigue detection method according to claim 6, characterized in that: Determining the corresponding red channel EEG energy feature, green channel EEG energy feature and blue channel EEG energy feature according to the θ wave energy feature, the δ wave energy feature and the α wave periodic energy feature corresponding to each signal collection position on the two-dimensional position specifically includes: The red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are respectively determined to be one of the following three types: the θ wave energy feature, the δ wave energy feature, the α wave periodic energy feature, and the red channel EEG energy feature, the green channel EEG energy feature and the blue channel EEG energy feature are different from each other.

11. The fatigue detection method according to claim 1, characterized in that: Also includes: The EEG information image is subjected to image recognition through a fatigue detection model to determine whether the EEG information image represents a wakeful state or a fatigued state.

12. The fatigue detection method according to claim 1, characterized in that: The detecting of the EEG signals corresponding to the multiple signal collection positions on the head specifically includes: Detecting the original EEG signals corresponding to multiple signal collection positions on the head; The raw EEG signal is preprocessed to obtain the EEG signal; wherein the preprocessing includes one or more of the following methods: filtering, removing power frequency interference, bad channel detection and interpolation, independent component analysis, re-referencing, data cleaning and screening.

13. A fatigue detection device, characterized in that: include: An EEG signal detection module is used to detect EEG signals corresponding to multiple signal collection positions on the head; A two-dimensional position determination module, used to determine the two-dimensional positions of the multiple signal collection positions according to the relative position relationship of the multiple signal collection positions; A pixel value determination module, used to determine the corresponding θ wave energy feature, δ wave energy feature and α wave periodic energy feature according to the EEG signals corresponding to each of the plurality of signal acquisition positions on the two-dimensional position, and determine the pixel value corresponding to each of the signal acquisition positions on the two-dimensional position; wherein the pixel value reflects the magnitude of the θ wave energy feature, δ wave energy feature and α wave periodic energy feature of the EEG signal corresponding to each of the signal acquisition positions on the two-dimensional position; The EEG information image determination module is used to determine the EEG information image according to the pixel value corresponding to each signal collection position on the two-dimensional position; the EEG information image is used for fatigue judgment.

14. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, and when the at least one instruction is executed by the processor, the fatigue detection method according to any one of claims 1 to 12 is implemented.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed, the fatigue detection method according to any one of claims 1 to 12 is implemented.

Images