Sensitive brain region and electroencephalogram characteristic determination method and device for fatigue detection

Abstract

The invention discloses a sensitive brain region and electroencephalogram characteristic determination method for fatigue detection. The determination method comprises the following steps: selecting a plurality of signal acquisition positions on the head and acquiring conscious electroencephalogram signals and fatigue electroencephalogram signals; determining various consciousness electroencephalogram characteristics and various fatigue electroencephalogram characteristics according to the consciousness electroencephalogram signals and the fatigue electroencephalogram signals acquired at each signal acquisition position; determining the difference degree of the conscious electroencephalogram features and the fatigue electroencephalogram features of the same type of each signal acquisition position, sorting the conscious electroencephalogram features and the fatigue electroencephalogram features to obtain a sorting result, and selecting a sensitive brain area and electroencephalogram features for fatigue judgment from the signal acquisition positions and the electroencephalogram features within a preset high difference degree range according to the sorting result. According to the method, the sensitive brain region and the electroencephalogram characteristics which are more beneficial to fatigue detection can be selected. The invention further discloses a sensitive brain region and electroencephalogram feature determination device for fatigue detection, a fatigue detection method, 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 method for determining sensitive brain areas and electroencephalogram characteristics for fatigue detection, a method device for determining sensitive brain areas and electroencephalogram characteristics for fatigue detection, a fatigue detection method, an electronic device and a computer-readable storage medium. Background Art

[0002] Due to long, irregular, night flights and cross-time zone flight changes, flight crews are prone to lack of sleep, fatigue and changes in biological rhythms, which seriously affect their work ability and easily cause flight safety hazards. The prevention and monitoring of flight fatigue is one of the important topics in current aerospace medical research. About 75% of aviation accidents are related to human errors, and the decline in work ability caused by flight fatigue is an important cause of accidents. Therefore, the establishment of a dynamic monitoring system for flight fatigue is of great significance to ensuring flight safety.

[0003] There are many methods for determining flight fatigue, such as scale assessment, objective observation, and physiological and biochemical methods, but the assessment system is not yet mature. Subjective methods such as fatigue assessment scales cannot objectively reflect the actual functional status of fatigued personnel. Objective observation methods include head position sensing, mouth state detection, pupil measurement, visual movement detection, and eyelid closure percentage detection. These technologies are based on long-distance transport aircraft pilots and civil aviation pilots, who have low loads and long-term mental consumption. They are not applicable to military pilots with high work intensity and high fatigue sensitivity.

[0004] EEG signals have the advantages of high temporal resolution and strong functional specificity, so they have become the gold standard for detecting fatigue based on physiological signals.

[0005] However, in the prior art, fatigue detection using EEG signals still has the following problems.

[0006] There is no unified indicator for fatigue detection through EEG signals. In 2009, a study used EEG signals to determine the driver's fatigue level. Using a simulated driving experiment, it was found that the larger the ratio of the EEG power spectrum (α+θ) / β, the higher the fatigue level; in 2013, researchers conducted an actual EEG characteristic analysis of the fatigue state of long-distance bus drivers. EEG collection was carried out during driving breaks, and it was concluded that the larger the θ / β, the higher the fatigue level; in 2021, researchers found that the Rayleigh entropy of EEG will continue to increase with the accumulation of fatigue; in 2021, researchers found that the prefrontal lobe is a sensitive brain area for fatigue, among which the brain functional connectivity index contributes the most to the fatigue recognition model. Therefore, it can be seen that there is no consensus and conclusion on the indicators and methods of fatigue detection.

[0007] There is a certain contradiction between the accuracy and ease of use of EEG fatigue detection. High-precision EEG fatigue detection usually requires complex equipment and multiple electrodes to capture the brain's electrical activity. For example, some studies have used EEG acquisition systems with up to 64 leads to synchronously collect the subjects' EEG signals. This complex setup can achieve high accuracy in a laboratory environment, but in practical applications, such as driving or in the workplace, it may reduce its ease of use due to the complexity and non-portability of the equipment.

[0008] In order to improve detection accuracy, researchers usually extract a large number of features from EEG signals and use complex algorithms for feature selection. For example, some studies have proposed using deep networks for multi-level feature generation, combining one-dimensional binary patterns (BP) and statistical features, and one-dimensional discrete wavelet transform (1D-DWT) to create multi-layer features. Although these advanced feature extraction and selection methods can improve accuracy, they also increase the complexity of the system and are not conducive to real-time and convenient use.

[0009] High-precision EEG fatigue detection often requires complex algorithms to process and analyze data. For example, convolutional neural networks (CNNs) are used to automatically extract features from EEG signals and classify their states. Although these algorithms perform well in terms of classification accuracy, they usually require high computing resources, which may limit their real-time applications on mobile or low-power devices.

[0010] In order to obtain high-quality EEG signals, users usually need to wear multiple electrodes, which may cause discomfort, especially when used for a long time. Some studies have proposed using forehead EEG signals to design a portable driver fatigue detection framework to reduce wearing discomfort and improve ease of use. However, reducing the number of electrodes may affect the quality of the signal and thus affect the accuracy of detection.

[0011] High-precision EEG fatigue detection devices are often expensive, which limits their popularity. Low-cost devices may not provide sufficient accuracy, thus limiting their effectiveness in practical applications.

[0012] It can be seen from this that how to select sensitive brain areas for fatigue detection, select EEG indicators with high sensitivity and good consistency, and make the number of selected sensitive brain areas and EEG indicators as small as possible and easy to calculate is a major difficulty. Summary of the invention

[0013] The present invention proposes a method for determining sensitive brain areas and EEG characteristics for fatigue detection to solve the above problems.

[0014] In a first aspect, an embodiment of the present invention discloses a method for determining a sensitive brain region and EEG features for fatigue detection, wherein the sensitive brain region is an area of ​​the head where EEG features are collected, and the determination method comprises:

[0015] 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 signals are EEG signals in the awake state, and the fatigue EEG signals are EEG signals in the fatigue state;

[0016] Determine, according to the awake EEG signals and fatigue EEG signals collected at each signal collection position, a plurality of awake EEG features corresponding to the awake EEG signals and a plurality of fatigue EEG features corresponding to the fatigue EEG signals;

[0017] 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, and select the final signal collection position and the final EEG feature as the sensitive brain area and the EEG feature for fatigue judgment from the signal collection positions and EEG features within a preset high difference degree range according to the sorting result; wherein the preset high difference degree range is a range with a relatively high degree of difference in the sorting.

[0018] By adopting the above technical scheme, the method for determining sensitive brain areas and EEG features for fatigue detection of the present invention can compare the degree of difference between the same type of awake EEG features and fatigue EEG features collected at different signal collection positions, and can also compare the sensitivity of different types of EEG features to fatigue, and then select the signal collection positions and EEG features within a preset high difference range by sorting, and then select sensitive brain areas and EEG features for fatigue judgment. It can accurately select sensitive brain areas and EEG features suitable for fatigue detection, and can also make the selected sensitive brain areas relatively small, and the selected EEG features relatively easier to calculate.

[0019] According to another specific embodiment of the present invention, the types of awake EEG features and fatigue EEG features include non-periodic offset features and non-periodic stretching and transformation features, and the non-periodic offset features and non-periodic stretching and transformation features of the EEG signal are determined by:

[0020] Determine the logarithmic power spectrum of the EEG signal;

[0021] According to the logarithmic power spectrum, nonlinear fitting is performed using the following formula 1 to obtain the corresponding fitting curve:

[0022] L(f)=ab×log(P(f))(Formula 1)

[0023] 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;

[0024] The offset parameter a in Formula 1 is determined to be a non-periodic offset feature, and the scaling parameter b in Formula 1 is determined to be a non-periodic scaling feature.

[0025] According to another specific embodiment of the present invention, the types of awake EEG features and fatigue EEG features also include α wave band energy features and α wave band peak features, and the α wave band energy features and α wave band peak features of the EEG signal are determined by:

[0026] The alpha wave spectrum periodic component is determined based on the logarithmic power spectrum and the fitting curve using the following formula 2:

[0027] P ′ (f) = P(f) - L(f) (Formula 2)

[0028] 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;

[0029] The α wave spectrum periodic energy component is determined by the following formula 3 based on the α wave spectrum periodic component:

[0030] P ″ (f) = 10 × 10 P′(f) (Formula 3)

[0031] 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;

[0032] The α wave frequency band energy characteristics and the α wave frequency band peak characteristics are determined according to the α wave spectrum period energy component.

[0033] According to another specific embodiment of the present invention, each of the multiple signal collection positions on the head collects the corresponding awake EEG signal and fatigue EEG signal, specifically including:

[0034] At each of the multiple signal collection positions on the head, a corresponding original awake EEG signal and an original fatigue EEG signal are respectively collected;

[0035] The original awake EEG signal and the original fatigue EEG signal are preprocessed to obtain awake EEG signal and fatigue EEG signal; wherein the preprocessing method includes at least one of the following: data segmentation, de-averaging, filtering, and windowing.

[0036] According to another specific embodiment of the present invention, the degree of difference between awake EEG features and fatigue EEG features of the same type is determined by a feature selection algorithm, and the final signal collection position and the final EEG feature are selected from the signal collection positions and EEG features within a preset high degree of difference range according to the sorting result, specifically including:

[0037] Determine the occurrence times of different signal acquisition positions and different EEG features within a preset high difference degree range;

[0038] The signal collection positions and EEG features with relatively large number of occurrences are selected as the position and feature candidate data sets corresponding to the feature selection algorithm, and the final signal collection position and final EEG features are selected from the position and feature candidate data sets.

[0039] According to another specific embodiment of the present invention, there are multiple feature selection algorithms, and the final signal collection position and the final EEG feature are selected from the signal collection positions and EEG features within a preset high difference range according to the sorting results, and further comprising:

[0040] Obtain the location and feature selection dataset corresponding to each feature selection algorithm;

[0041] In the union of the plurality of position and feature candidate data sets, the signal collection position and the EEG feature with relatively more occurrences are selected as the final signal collection position and the final EEG feature.

[0042] According to another specific embodiment of the present invention, the multiple feature selection algorithms include a feature selection algorithm based on machine learning and a feature selection algorithm based on statistics.

[0043] According to another specific embodiment of the present invention, multiple feature selection algorithms include minimum redundancy maximum relevance MRMR algorithm, chi-square test algorithm, ReliefF algorithm, Kruskal-Wallis test algorithm, and recursive feature sorting algorithm SVM-REF based on support vector machine.

[0044] According to another specific embodiment of the present invention, selecting the final signal collection position and the final EEG feature from the signal collection positions and EEG features within a preset high difference range according to the sorting result specifically includes:

[0045] Obtain a fatigue detection model, which detects fatigue based on signal acquisition position and EEG features;

[0046] Selecting a pending signal collection position and a pending EEG feature from within a preset high difference degree range, inputting the pending signal collection position and the pending EEG feature into a fatigue detection model and performing fatigue detection, and determining the accuracy of fatigue judgment by the fatigue detection model;

[0047] The number and / or type of pending signal collection positions and pending EEG features are changed, and the changed pending signal collection positions and pending EEG features are input into the fatigue detection model to determine the accuracy of fatigue judgment, until the minimum number of pending signal collection positions and pending EEG features whose determined accuracy meets the preset high accuracy conditions are obtained, and the pending signal collection positions and pending EEG features corresponding to the last change are determined as the final signal collection positions and final EEG features.

[0048] In a second aspect, an embodiment of the present invention discloses a device for determining a sensitive brain region and EEG characteristics for fatigue detection, wherein the sensitive brain region is an area of ​​the head where EEG characteristics are collected, and the device for determining comprises:

[0049] A signal acquisition module, used to select multiple signal acquisition positions on the head and acquire awake EEG signals and fatigue EEG signals corresponding to each signal acquisition position; wherein the awake EEG signals are EEG signals in the awake state, and the fatigue EEG signals are EEG signals in the fatigue state;

[0050] A feature calculation module, for determining a plurality of awake EEG features corresponding to the awake EEG signals and a plurality of fatigue EEG features corresponding to the fatigue EEG signals according to the awake EEG signals and fatigue EEG signals collected at each signal collection position;

[0051] The sensitive brain area and EEG feature determination module is used to 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 the sorting results, and select the final signal collection positions and the final EEG features as the sensitive brain areas and the EEG features for fatigue judgment from the signal collection positions and EEG features within a preset high difference degree range according to the sorting results; wherein the preset high difference degree range is a range with a relatively high degree of difference in the sorting.

[0052] By adopting the above technical scheme, the device for determining sensitive brain areas and EEG features for fatigue detection of the present invention can compare the degree of difference between the same type of awake EEG features and fatigue EEG features collected at different signal collection positions, and can also compare the sensitivity of different types of EEG features to fatigue, and then select the signal collection positions and EEG features within a preset high difference range by sorting, and then select sensitive brain areas and EEG features for fatigue judgment. It can accurately select sensitive brain areas and EEG features suitable for fatigue detection, and can also make the selected sensitive brain areas relatively small, and the selected EEG features relatively easier to calculate.

[0053] In a third aspect, an embodiment of the present invention discloses a fatigue detection method, which performs fatigue detection using the sensitive brain areas and EEG features determined by the method for determining sensitive brain areas and EEG features for fatigue detection in any of the aforementioned embodiments.

[0054] By adopting the above technical solution, the fatigue detection method of the present invention can achieve accurate fatigue detection through as few sensitive brain areas and EEG features as possible, and can reduce the amount of calculation.

[0055] According to another specific embodiment of the present invention, the determined sensitive brain areas include the F3, F4, O1, O2, and Oz electrode positions of the 10-20 international standard lead system, and the determined EEG characteristics include delta wave energy characteristics, theta wave energy characteristics, alpha wave energy characteristics, alpha wave peak characteristics, beta wave energy characteristics, non-periodic offset characteristics, non-periodic stretching and transformation characteristics, alpha wave frequency band energy characteristics, and alpha wave frequency band peak characteristics.

[0056] In a fourth 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, implements the method for determining sensitive brain areas and EEG characteristics for fatigue detection in any of the aforementioned embodiments, and / or implements the fatigue detection method in any of the aforementioned embodiments.

[0057] By adopting the above technical solution, the electronic device can compare the difference between the same type of awake EEG features and fatigue EEG features collected at different signal collection positions, and can also compare the sensitivity of different types of EEG features to fatigue, and then select the signal collection positions and EEG features within the preset high difference range by sorting, and then select sensitive brain areas and EEG features for fatigue judgment, which can accurately select sensitive brain areas and EEG features suitable for fatigue detection, and can also make the selected sensitive brain areas relatively small, and the selected EEG features are relatively easier to calculate. In addition, when performing fatigue detection, accurate fatigue detection can be achieved with as few sensitive brain areas and EEG features as possible, and the amount of calculation can be reduced.

[0058] In a fifth 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, it implements the method for determining sensitive brain areas and EEG characteristics for fatigue detection in any embodiment, and / or implements the fatigue detection method in any of the aforementioned embodiments.

[0059] By adopting the above technical solution, the computer-readable storage medium can compare the degree of difference between the same type of awake EEG features and fatigue EEG features collected at different signal collection positions, and can also compare the sensitivity of different types of EEG features to fatigue, and then select the signal collection positions and EEG features within the preset high difference range by sorting, and then select sensitive brain areas and EEG features for fatigue judgment, which can accurately select sensitive brain areas and EEG features suitable for fatigue detection, and can also make the selected sensitive brain areas relatively small, and the selected EEG features are relatively easier to calculate. In addition, when performing fatigue detection, accurate fatigue detection can be achieved with as few sensitive brain areas and EEG features as possible, and the amount of calculation can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 Schematic diagram of a process for determining sensitive brain regions and EEG features for fatigue detection in an embodiment of the present invention Figure 1 ;

[0061] Figure 2 Schematic diagram of a process for determining sensitive brain regions and EEG features for fatigue detection in an embodiment of the present invention Figure 2 ;

[0062] Figure 3 Schematic diagram of a process for determining sensitive brain regions and EEG features for fatigue detection in an embodiment of the present invention Figure 3 ;

[0063] Figure 4 A schematic diagram showing the structure of a method and device for determining sensitive brain areas and EEG characteristics for fatigue detection in an embodiment of the present invention;

[0064] Figure 5 A schematic diagram showing the structure of an electronic device in an embodiment of the present invention is shown;

[0065] Figure 6 A schematic diagram showing a logarithmic power spectrum and a fitting curve in an embodiment of the present invention;

[0066] Figure 7 A schematic diagram showing the periodic energy component of the alpha wave spectrum in an embodiment of the present invention;

[0067] Fig. 8A A brain topography map showing the statistical p-value corresponding to the delta wave energy characteristics in an embodiment of the present invention;

[0068] Figure 8B A brain topographic map showing the statistical t value corresponding to the delta wave energy characteristic in an embodiment of the present invention;

[0069] Fig.9A A brain topography map showing the statistical p-value corresponding to the energy characteristics of theta waves in an embodiment of the present invention;

[0070] Fig. 9B A brain topographic map showing the statistical t value corresponding to the energy characteristics of theta waves in an embodiment of the present invention;

[0071] Fig. 10A A brain topography map showing the statistical p-value corresponding to the alpha wave energy characteristics in an embodiment of the present invention;

[0072] Fig. 10B A brain topographic map showing the statistical t value corresponding to the alpha wave energy characteristic in an embodiment of the present invention;

[0073] Fig.11A A brain topographic map showing the statistical p value corresponding to the alpha wave peak energy characteristic in an embodiment of the present invention;

[0074] Fig. 11B A brain topographic map showing the statistical t value corresponding to the α wave peak energy characteristic in an embodiment of the present invention;

[0075] Fig. 12A A brain topographic map showing the statistical p-value corresponding to the beta wave energy characteristics in an embodiment of the present invention;

[0076] Fig. 12B A brain topographic map showing the statistical t value corresponding to the beta wave energy characteristic in an embodiment of the present invention;

[0077] Fig.13A A brain topography map showing the statistical p-value corresponding to the non-periodic telescopic transformation feature in an embodiment of the present invention;

[0078] Fig. 13B A brain topographic map showing the statistical t value corresponding to the non-periodic telescopic transformation feature in an embodiment of the present invention;

[0079] Fig.14A A brain topography map showing the statistical p-value corresponding to the non-periodic offset feature in an embodiment of the present invention;

[0080] Fig. 14B A brain topography map showing the statistical t value corresponding to the non-periodic offset feature in an embodiment of the present invention;

[0081] Fig.15A A brain topography map showing the statistical p-value corresponding to the alpha wave frequency band energy characteristics in an embodiment of the present invention;

[0082] Fig. 15BA brain topographic map showing the statistical t value corresponding to the alpha wave frequency band energy characteristics in an embodiment of the present invention;

[0083] Fig.16A A brain topography map showing the statistical p value corresponding to the peak characteristics of the alpha wave frequency band in an embodiment of the present invention;

[0084] Fig. 16B A brain topographic map showing the statistical t value corresponding to the peak characteristics of the α wave frequency band in an embodiment of the present invention;

[0085] Fig.17 A statistical schematic diagram showing the occurrence times of different EEG features in an embodiment of the present invention;

[0086] Fig.18 The figure shows the brain topography of sensitive brain areas obtained by using the chi-square test algorithm in an embodiment of the present invention;

[0087] Fig.19 The brain topography of sensitive brain areas obtained by using the minimum redundancy maximum correlation (MRMR) algorithm in an embodiment of the present invention is shown;

[0088] Fig. 20 The brain topography of sensitive brain areas obtained by using the Kruskal-Wallis test algorithm in an embodiment of the present invention is shown;

[0089] Fig.21 The brain topography of sensitive brain areas obtained by using the ReliefF algorithm in an embodiment of the present invention is shown;

[0090] Fig. 22 The brain topography of sensitive brain areas obtained by using a recursive feature ranking algorithm based on a support vector machine (SVM-REF) in an embodiment of the present invention is shown;

[0091] Fig.23 The brain topography map of sensitive brain areas obtained by integrating multiple feature selection algorithms in an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0092] 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.

[0093] 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.

[0094] The terms “first”, “second”, etc. are only used for distinguishing descriptions and should not be understood as indicating or implying relative importance.

[0095] 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.

[0096] In order 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.

[0097] Electroencephalogram (EEG) fatigue detection is a technology that uses EEG signals to assess an individual's fatigue state. The basis of this technology is that EEG waves can reflect the electrical activity of the brain, and these activity patterns will change with the changes in the brain's fatigue state. Brain waves are usually divided into several major bands, including 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. In EEG fatigue detection, researchers analyze the time domain characteristics (such as amplitude, power spectral density) and frequency domain characteristics (such as frequency band power) of brain waves to identify changes in EEG patterns that are unique to normal sleep, wakefulness, and fatigue stages.

[0098] When the prior art detects fatigue through brain waves, it usually requires complex equipment to collect EEG signals from multiple locations on the head, and the number of locations may be as many as 64, for example. There are many types of EEG features corresponding to EEG signals. In order to achieve high-precision fatigue detection, the prior art also processes EEG data and calculates EEG features through complex calculation methods. This fatigue detection method of the prior art collects a large amount of data, a large amount of calculation, high cost, and low popularity. It is not suitable for use during the flight of pilots, and it is difficult to meet the needs of pilots for fast and accurate fatigue detection during the flight. Therefore, it is necessary to further reduce the number of locations on the head where EEG signals are collected, reduce or simplify the EEG indicators that need to be calculated, and at the same time ensure high-precision fatigue detection, so that fatigue detection can be performed during the flight of pilots to improve flight safety.

[0099] First, refer to Figure 1 The embodiment of the present invention discloses a method for determining a sensitive brain region and EEG features for fatigue detection, wherein the sensitive brain region is an area of ​​the head where EEG features are collected, and the determination method comprises the following steps:

[0100] S1: Select multiple signal collection positions on the head and collect awake EEG signals and fatigue EEG signals corresponding to each signal collection position.

[0101] Among them, 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.

[0102] S2: determining a plurality of awake EEG features corresponding to the awake EEG signals and a plurality of fatigue EEG features corresponding to the fatigue EEG signals according to the awake EEG signals and the fatigue EEG signals collected at each signal collection position.

[0103] S3: 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 the sorting result, and select the final signal collection position and the final EEG feature from the signal collection positions and EEG features within a preset high difference range as the sensitive brain area and the EEG feature for fatigue judgment according to the sorting result.

[0104] The preset high difference degree range is a range with relatively high difference degrees in the sorting.

[0105] Relatively high more specifically means that in the sorting results, the degree of difference between the same type of awake EEG features and fatigue EEG features at a specific signal acquisition position is higher than the degree of difference between the same type of awake EEG features and fatigue EEG features at other signal acquisition positions; or it may mean that among multiple types of awake EEG features and fatigue EEG features at a specific signal acquisition position, the degree of difference between a certain type of awake EEG features and fatigue EEG features is higher than other types of awake EEG features and fatigue EEG features; or it may also mean that the degree of difference between the same type of awake EEG features and fatigue EEG features at a specific signal acquisition position is higher than the degree of difference between other types of awake EEG features and fatigue EEG features at other signal acquisition positions. The degree of difference between the awake EEG features and fatigue EEG features corresponding to the signal acquisition positions and EEG features within the preset high difference degree range is higher than that of other signal acquisition positions and other EEG features.

[0106] Specifically, the preset high difference degree range may be, for example, the first 50% of the sorting results. Further, the preset high difference degree range may be, for example, the first 20%-first 30% of the sorting results. Still further, the preset high difference degree range may be, for example, the first 5%-first 10% of the sorting results. The higher the preset high difference degree range is in the entire sorting results, the more conducive it is to the subsequent selection of EEG features and sensitive brain areas that are more sensitive to fatigue.

[0107] By adopting the above technical scheme, the method for determining sensitive brain regions and EEG features for fatigue detection of the present invention can compare the degree of difference between the same type of awake EEG features and fatigue EEG features collected at different signal collection positions, and can also compare the sensitivity of different types of EEG features to fatigue, and then select the signal collection positions and EEG features within the preset high difference range by sorting, and then select sensitive brain regions and EEG features for fatigue judgment, which can accurately select sensitive brain regions and EEG features suitable for fatigue detection, and can also make the selected sensitive brain regions relatively small, and the selected EEG features are relatively easier to calculate. Compared with the prior art method of detecting EEG signals at multiple positions (for example, 64 positions) of the head and making fatigue judgments, the above method of the present invention can select a relatively small number of final signal collection positions and relatively small types of final EEG features, and can exclude most of the signal collection positions and EEG features that have low contribution to fatigue detection and are basically irrelevant to fatigue, thereby reducing the amount of calculation.

[0108] Furthermore, there may be multiple final signal collection positions and multiple final EEG features.

[0109] In the above embodiment, the types of awake EEG features and fatigue EEG features may include δ wave energy features, θ wave energy features, α wave energy features, α wave peak features, and β wave energy features. The above five features can all be calculated by obtaining a power spectrum after performing digital discrete Fourier transform on the EEG signal.

[0110] The inventor has found through research that non-periodic features in EEG features can also be used to judge fatigue. Non-periodic features include, for example, non-periodic offset features and non-periodic scaling features. In the above embodiment, the types of awake EEG features and fatigue EEG features include non-periodic offset features and non-periodic scaling features. The non-periodic offset features and non-periodic scaling features of EEG signals are determined in the following ways:

[0111] S01: Determine the logarithmic power spectrum of the EEG signal.

[0112] refer to Figure 6 , Figure 6 Curve 2 in is the logarithmic power spectrum of the EEG signal. Specifically, the logarithmic power spectrum of the EEG signal can be calculated by the following formula 4:

[0113]

[0114] Wherein, y is the power spectrum obtained after performing digital discrete Fourier transform on the EEG signal, L is the length of the digital discrete Fourier transform, and F is the logarithmic power spectrum. Wherein, the length of the digital discrete Fourier transform can be consistent with the length of the EEG signal.

[0115] S02: Perform nonlinear fitting based on the logarithmic power spectrum using the following formula 1 to obtain the corresponding fitting curve:

[0116] L(f)=ab×log(P(f))(Formula 1)

[0117] 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.

[0118] refer to Figure 6 , in this embodiment, Figure 6 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.

[0119] S03: Determine that the offset parameter a in Formula 1 is a non-periodic offset feature, and determine that the scaling parameter b in Formula 1 is a non-periodic scaling feature.

[0120] In the present invention, the non-periodic offset characteristics and non-periodic stretching and transformation characteristics calculated by curve fitting using the above method and formula have a relatively high sensitivity to fatigue, which helps to improve the accuracy of fatigue detection.

[0121] Furthermore, the inventors found that the α wave energy characteristics and α wave peak characteristics often used in the prior art are not sensitive enough to fatigue. In response to this, the inventors further improved the EEG characteristics corresponding to the α wave selected during fatigue detection. The inventors found that the α wave frequency band energy characteristics and α wave frequency band peak characteristics are more sensitive to fatigue.

[0122] In the above embodiment, the types of awake EEG features and fatigue EEG features also include α wave band energy features and α wave band peak features. The α wave band energy features and α wave band peak features of the EEG signal are determined by:

[0123] S04: Determine the alpha wave spectrum periodic component according to the logarithmic power spectrum and the fitting curve using the following formula 2:

[0124] P ′ (f) = P(f) - L(f) (Formula 2)

[0125] 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.

[0126] S05: Determine the α wave spectrum periodic energy component according to the α wave spectrum periodic component by the following formula 3:

[0127] P ″ (f) = 10 × 10 P′(f) (Formula 3)

[0128] 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. Figure 7 , Figure 7 Curve 3 in the figure is the periodic energy component of the α wave spectrum.

[0129] S06: Determine the α wave frequency band energy characteristics and the α wave frequency band peak characteristics according to the α wave spectrum period energy component.

[0130] The α wave spectrum periodic energy component P can be obtained by the above formula 3. ″ (f), α wave spectrum periodic energy component P″ (f) dimensionless, which can then be calculated based on P ″ (f) Calculate the energy corresponding to the α band, that is, the α band energy characteristics, and also obtain the α band peak characteristics. The α band energy characteristics can correspond to the α wave spectrum periodic energy component P ″ (f) The area enclosed by the coordinate axis in the α band (i.e., 8-13 Hz). The peak characteristics of the α band can correspond to the periodic energy component P of the α wave spectrum. ″ (f) The peak value.

[0131] The α wave frequency band energy characteristics and α wave frequency band peak characteristics calculated by the above method and formula are more sensitive to fatigue and help improve the accuracy of fatigue detection.

[0132] Further, refer to Figure 2 , the aforementioned step S1 specifically includes the following steps:

[0133] S11: respectively collecting corresponding original awake EEG signals and original fatigue EEG signals at each of the multiple signal collection positions on the head.

[0134] Specifically, the brain waves of the subject in a sober state and in a fatigued state can be collected respectively, so as to obtain original sober EEG signals and original fatigue EEG signals respectively.

[0135] S12: preprocessing the original awake EEG signal and the original fatigue EEG signal to obtain awake EEG signal and fatigue EEG signal. The preprocessing method includes at least one of the following: data segmentation, de-averaging, filtering, and windowing.

[0136] Specifically, data segmentation refers to segmenting the collected EEG signals to increase the amount of data and improve the comprehensiveness of the data, so as to improve the versatility of the subsequently selected sensitive brain areas and EEG features for fatigue detection, and to facilitate the subsequent verification of whether the selected sensitive brain areas and EEG features for fatigue detection are effective, which is conducive to the subsequent construction of a more stable and accurate fatigue detection model. Exemplarily, the EEG signal can be sliced ​​step by step according to a certain step size. For example, for an EEG signal with a duration of 3 minutes, each segment of the EEG signal can be 4-10 seconds, and the step can be 1-5 seconds.

[0137] De-averaging can be used to eliminate the baseline, perform baseline correction, and improve the accuracy of subsequent analysis of EEG signals.

[0138] The filtering can adopt the traditional FIR filtering (Finite Impulse Response, finite length unit impulse response filtering), which can maintain the stability of phase change, and the bandpass range is f1-f2Hz. Among them, the range of f1 is 0.5-2Hz, and the low frequency band in the brain wave should be retained as much as possible, such as the δ band (0.5-4Hz); the range of f2 is 30-45z, and the high frequency band in the brain wave should be retained as much as possible, such as the β band (14-30Hz). Exemplarily, the bandpass range can be 1-45Hz.

[0139] Windowing refers to adding a window to the EEG signal, which can reduce the impact of endpoint data on the spectrum measurement of subsequent signals. You can choose to add a Hamming window or a Hanning window.

[0140] Among them, adopting one of the above-mentioned multiple preprocessing methods is conducive to the subsequent analysis of the EEG signal. When all of the above-mentioned multiple preprocessing methods are adopted, a better effect can be achieved and the accuracy of the analysis of the EEG signal can be further improved.

[0141] Furthermore, the degree of difference between the awake EEG features and the fatigue EEG features of the same type is determined by a feature selection algorithm. Analyzing the degree of difference between the awake EEG features and the fatigue EEG features by a feature selection algorithm helps to find out the type of EEG features that contribute more to fatigue detection.

[0142] refer to Figure 3 In the aforementioned step S3, the final signal collection position and the final EEG feature are selected from the signal collection positions and EEG features within the preset high difference degree range according to the sorting result, which specifically includes the following steps:

[0143] S31: Determine the number of occurrences of different signal collection positions and different EEG features within a preset high difference degree range.

[0144] S32: Select signal acquisition positions and EEG features that appear relatively frequently as position and feature candidate data sets corresponding to the feature selection algorithm, and select the final signal acquisition position and final EEG feature from the position and feature candidate data sets.

[0145] Among them, the signal collection positions and EEG features that appear relatively more times within the preset high difference degree range can be regarded as signal collection positions and EEG features that are relatively more sensitive to fatigue, which helps to improve the accuracy of fatigue detection.

[0146] More specifically, a relatively large number of occurrences may mean that within a preset high difference degree range, a specific signal acquisition position and a specific type of EEG feature appear more frequently than other signal acquisition positions and other types of EEG features.

[0147] Exemplarily, the signal acquisition positions and EEG features can be sorted according to the number of occurrences to obtain the occurrence sorting result. The signal acquisition positions and EEG features with relatively more occurrences can refer to the signal acquisition positions and EEG features that are within the preset multiple occurrence range in the occurrence sorting result. The signal acquisition positions and EEG features within the preset multiple occurrence range have more occurrences than other signal acquisition positions and other EEG features. Among them, the preset multiple occurrence range can be, for example, the top 80% in the occurrence sorting result. Further, the preset multiple occurrence range can be, for example, the top 40%-top 60% in the occurrence sorting result. Further, the preset multiple occurrence range can be, for example, the top 10%-top 25% in the occurrence sorting result. The higher the preset multiple occurrence range is in the entire occurrence sorting result, the more conducive it is to the subsequent selection of EEG features and sensitive brain areas that are more sensitive to fatigue.

[0148] Furthermore, in the above embodiments, there are multiple feature selection algorithms. In the above step S3, the final signal collection position and the final EEG feature are selected from the signal collection positions and EEG features within the preset high difference range according to the sorting result, and further includes:

[0149] S33: Obtain the position and feature selection dataset corresponding to each feature selection algorithm.

[0150] Specifically, each feature selection algorithm corresponds to a position and a feature selection dataset, and multiple feature selection algorithms can obtain multiple positions and feature selection datasets.

[0151] S34: From the union of the plurality of position and feature data sets to be selected, select the signal collection position and the EEG feature with a relatively large number of occurrences as the final signal collection position and the final EEG feature.

[0152] When multiple feature selection algorithms are used, there are also multiple positions and feature data sets to be selected, and the union of multiple positions and feature data sets to be selected can be obtained accordingly. Accordingly, the final signal acquisition position and the final EEG feature can be selected from the union of multiple positions and feature data sets to be selected. Among them, the signal acquisition position and EEG feature that appear relatively more times in the union of multiple positions and feature data sets to be selected can be selected as the final signal acquisition position and the final EEG feature. Among them, the union refers to the sum of multiple positions and feature data sets to be selected, and the union is composed of all elements in multiple positions and feature data sets to be selected, that is, in the union, the number of occurrences of each signal acquisition position is the sum of the number of times each signal acquisition position appears in multiple positions and feature data sets to be selected, and the number of occurrences of each EEG feature is the sum of the number of times each EEG feature appears in multiple positions and feature data sets to be selected.

[0153] Exemplarily, in the union of multiple positions and feature selection data sets, the signal acquisition positions and EEG features can be sorted again according to the number of occurrences to obtain a corresponding occurrence sorting result, which can be recorded as a second occurrence sorting result, for example. At this time, the signal acquisition positions and EEG features with relatively more occurrences can refer to the signal acquisition positions and EEG features that are located in the second preset multiple occurrence range in the second occurrence sorting result. The signal acquisition positions and EEG features located in the second preset multiple occurrence range have more occurrences than other signal acquisition positions and other EEG features in the union of multiple positions and feature selection data sets. Among them, the second preset multiple occurrence range can be, for example, the first 80% in the second occurrence sorting result. Further, the second preset multiple occurrence range can be, for example, the first 40%-first 60% in the second occurrence sorting result. Further, the second preset multiple occurrence range can be, for example, the first 10%-first 25% in the second occurrence sorting result. The closer the second preset multiple occurrence range is in the entire second occurrence sorting result, the more conducive it is to the subsequent selection of EEG features and sensitive brain areas that are more sensitive to fatigue.

[0154] Furthermore, the signal collection position and EEG feature with the largest number of occurrences can be selected from the position and feature candidate data sets as the final signal collection position and final EEG feature, or the signal collection position and EEG feature with the largest number of occurrences can be selected from the union of multiple position and feature candidate data sets as the final signal collection position and final EEG feature.

[0155] When a specific signal collection position and a specific EEG feature appear relatively frequently in the union of multiple positions and feature candidate data sets, it means that when the difference between wakefulness and fatigue is calculated and sorted by different feature selection algorithms, the specific signal collection position and the specific EEG feature are ranked relatively high, appear relatively frequently, and the difference is highly stable, which can be explained as being more sensitive to fatigue. The final signal collection position and the final EEG feature selected in the above manner are more sensitive to fatigue, while ensuring that the number is relatively small.

[0156] Furthermore, multiple feature selection algorithms include feature selection algorithms based on machine learning and feature selection algorithms based on statistics. By using feature selection methods from various fields including machine learning and statistics to evaluate and rank the differences between awake EEG features and fatigue EEG features, the evaluation method is more comprehensive, the evaluation results obtained are more accurate, the results are more stable, and it helps to determine a more reliable final signal acquisition position and final EEG features, with higher versatility.

[0157] Furthermore, in the above embodiment, the number of feature selection algorithms may be, for example, 3-10. Preferably, the number of feature selection algorithms may be 4-7. Such a setting can ensure the comprehensiveness of the evaluation method without causing too many evaluation methods and too much calculation, and can select EEG features and signal acquisition positions that are highly versatile, highly reliable, and more sensitive to fatigue.

[0158] Exemplarily, the plurality of feature selection algorithms include a minimum redundancy maximum relevance (MRMR) algorithm, a chi-square test algorithm, a ReliefF algorithm, a Kruskal-Wallis test algorithm, and a support vector machine-based recursive feature ranking algorithm (SVM-REF).

[0159] Among them, the minimum redundancy maximum relevance (MRMR) algorithm can be recorded as the fscmrmr method. The MRMR algorithm aims to find a feature subset from the feature set that has the maximum relevance (Max-Relevance) to the target variable and the minimum redundancy (Min-Redundancy) between features.

[0160] The chi-square test algorithm can be recorded as the fscchi2 method. The chi-square test can predict the variables that are most influential on the classification result. In this embodiment, the classification result can be divided into two categories, such as awake and tired.

[0161] The ReliefF algorithm, denoted as the relieff method, ranks the importance of predictive variables. The ReliefF algorithm is based on the following idea: good features should be able to reduce the distance between samples of the same type and increase the distance between samples of different types.

[0162] The Kruskal-Wallis test algorithm is recorded as the Kruskalwallis method. Kruskal-Wallis is a nonparametric statistical method used to determine whether there is a statistically significant difference between the medians of three or more independent groups. The Kruskal-Wallis test can be performed on the data using statistical software or programming languages ​​such as Python or R. The test results usually include an H value (Kruskal-Wallis statistic) and a p-value. The rank of each sample is calculated during the Kruskal-Wallis test. The rank is the ordinal position of the sample values ​​in all merged data sets, and can be sorted according to the rank.

[0163] The recursive feature ranking algorithm based on support vector machine (SVM-REF), which can be denoted as SVMRFECBR, SVM-RFE (Recursive Feature Elimination with Support Vector Machines) is a feature selection method that reduces the number of features by recursively removing the features that have the least impact on the model.

[0164] Further, in the above embodiments, in the above step S3, selecting the final signal collection position and the final EEG feature from the signal collection positions and EEG features within the preset high difference degree range according to the sorting result may specifically include the following steps:

[0165] S35: Obtain a fatigue detection model, which performs fatigue detection through signal acquisition position and EEG features.

[0166] Exemplarily, the following machine learning models can be used to build a fatigue detection model: Fine Tree, Medium Tree, Coarse Tree, Linear Discriminant, Quadratic Discriminant, Binary GLM Logistic Regression, Efficient Logistic Regression, Efficient Linear SVM, Gaussian Naive Bayes, Kernel Naive Bayes, Linear SVM, Quadratic SVM, Cubic SVM, Fine Gaussian SVM, Medium Gaussian SVM, Coarse Gaussian SVM, Fine KNN, Medium KNN, etc. KNN), Coarse KNN, Cosine KNN, Cubic KNN, Weighted KNN, Boosted Trees, Bagged Trees, Subspace Discriminant Analysis, Subspace KNN, RUS Boosted Trees (Random Under-Sampling Boosted Trees, which can be abbreviated as RUSBoosted Trees), Narrow Neural Network, Medium Neural Network, Wide Neural Network, Bilayered Neural Network, Trilayered Neural Network, SVM Kernel, Logistic Regression Kernel. Other types of machine learning models can also be used to construct fatigue detection models, and the present invention does not limit this. It only needs to satisfy that the fatigue detection model can perform fatigue detection through signal acquisition position and EEG characteristics.

[0167] S36: Select a pending signal collection position and a pending EEG feature from within a preset high difference degree range, input the pending signal collection position and the pending EEG feature into a fatigue detection model and perform fatigue detection, and determine the accuracy of fatigue judgment by the fatigue detection model.

[0168] Exemplarily, the accuracy can be determined by calculating the confusion matrix of the fatigue detection model, and the confusion matrix can be used to evaluate the performance of the fatigue detection model. More specifically, the accuracy of the fatigue detection model on the test set can be calculated. The test set consists of awake EEG signals and fatigued EEG signals. Random EEG signals in the test set are input into the fatigue detection model and judged to obtain a judgment result. The judgment result is compared with the actual state corresponding to the EEG signal of the input model, and the accuracy of the fatigue detection model in fatigue judgment can be obtained. The accuracy rate is more specifically the proportion of the number of data correctly judged by the model to the total number of data.

[0169] S37: Change the number and / or type of pending signal collection positions and pending EEG features, and input the changed pending signal collection positions and pending EEG features into the fatigue detection model to determine the accuracy of fatigue judgment, until the minimum number of pending signal collection positions and pending EEG features whose determined accuracy meets the preset high accuracy conditions are obtained, and the pending signal collection positions and pending EEG features corresponding to the last change are determined as the final signal collection positions and final EEG features.

[0170] The number and / or type of pending signal collection positions and pending EEG features may be updated based on the signal collection positions and EEG features within a preset high difference degree range.

[0171] Furthermore, when a feature selection algorithm is used to sort the degree of difference, the number and / or type of pending signal collection positions and pending EEG features can also be updated based on the positions corresponding to the feature selection algorithm and the signal collection positions and EEG features in the feature candidate data set. Furthermore, when multiple feature selection algorithms are used, the number and / or type of pending signal collection positions and pending EEG features can also be updated based on the signal collection positions and EEG features in the multiple positions and feature candidate data set, or the number and / or type of pending signal collection positions and pending EEG features can be updated based on the signal collection positions and EEG features in the intersection of multiple positions and feature candidate data sets.

[0172] Exemplarily, changing the number and / or type of pending signal acquisition positions and pending EEG features may specifically include gradually reducing the number of pending signal acquisition positions or gradually reducing the types of EEG features according to the sorting results obtained in the aforementioned method. When changing the number and / or type of pending signal acquisition positions and pending EEG features, priority is given to and retention of pending signal acquisition positions and pending EEG features with higher accuracy. Specifically, when it is necessary to reduce the number of pending signal acquisition positions or gradually reduce the types of EEG features, the signal acquisition positions and EEG features that are relatively low in ranking in the sorting results or appear relatively few times within the preset high difference range can be preferentially removed, that is, the signal acquisition positions and EEG features with low fatigue correlation can be preferentially excluded. In a possible embodiment, if the difference in accuracy re-determined after removing only the pending signal acquisition position A and removing only the pending signal acquisition position B is small, the signal acquisition position with higher accuracy re-determined is preferentially retained, and the signal acquisition position with lower accuracy re-determined is removed. Furthermore, the number of signal collection positions can be changed preferentially while the EEG features remain unchanged. After the final signal collection positions are determined, a similar method is used to gradually reduce the types of calculated EEG features. That is, when removing EEG features, the types of EEG features with low correlation are removed preferentially, and when replacing EEG features, the types of EEG features with higher accuracy that are re-determined are retained preferentially.

[0173] Similarly, changing the number and / or type of pending signal acquisition positions and pending EEG features may specifically include gradually increasing the number of pending signal acquisition positions according to the sorting results obtained in the aforementioned method, or gradually increasing the types of calculated EEG features. When changing the number and / or type of pending signal acquisition positions and pending EEG features, priority is given to pending signal acquisition positions and pending EEG features with higher accuracy. Specifically, when it is necessary to increase the number of pending signal acquisition positions or gradually increase the types of EEG features, priority can be given to increasing signal acquisition positions and EEG features that are relatively high in the sorting results or appear relatively frequently within a preset high difference range, that is, priority can be given to increasing signal acquisition positions and EEG features with high fatigue relevance. In a possible embodiment, if the difference in accuracy re-determined after only adding the pending signal acquisition position A and only adding the pending signal acquisition position B is small, priority can be given to increasing the re-determined signal acquisition position with higher accuracy. Furthermore, the number of signal collection positions can be changed preferentially while the EEG features remain unchanged. After the final signal collection positions are determined, a similar method is used to gradually increase or replace the types of calculated EEG features. That is, when adding EEG features, the types of EEG features with high correlation are increased preferentially, and when replacing EEG features, the types of EEG features that have been re-determined with higher accuracy are retained preferentially.

[0174] The preset high accuracy rate condition may be, for example, a set accuracy rate threshold. When the determined accuracy rate is greater than or equal to the accuracy rate threshold, it may be considered that the preset high accuracy rate condition is met. Exemplarily, the accuracy rate threshold may be, for example, 0.75-0.95. Further, the accuracy rate threshold may also be, for example, 0.80-0.90.

[0175] Furthermore, the number and / or type of signal acquisition positions can be adjusted preferentially, while the number and type of EEG features remain unchanged. For example, the pending signal acquisition positions can be increased, decreased or replaced preferentially, and all nine of the above-mentioned pending EEG features are selected to determine the accuracy. When the determined accuracy meets the preset high accuracy condition, the pending signal acquisition position corresponding to the last adjustment can be determined as the final signal acquisition position. After determining the final signal acquisition position, the type and / or number of EEG features can be adjusted. For example, the type of EEG features can be gradually reduced or replaced to minimize the number of calculated EEG features while ensuring high accuracy. Using the above method, it is possible to determine EEG features and signal acquisition positions with as few as possible numbers, as little calculation as possible, and high fatigue detection accuracy.

[0176] A more specific embodiment is described below.

[0177] 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.

[0178] 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.

[0179] 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', 'F1', 'F3', 'F5', 'F7', 'FC1', 'FC3', 'FC5', 'FT7', 'Cz', 'C1', 'C3', 'C5', 'T7', 'CP1', 'CP3', 'CP5', 'TP7', 'TP9', 'Pz', 'P1', 'P3', 'P5', 'P7', 'P O3', 'PO7', 'Oz', 'O1', 'Fpz', 'Fp2', 'AF4', 'AF8', 'F2', 'F4', 'F6', 'F8', 'FC2', 'FC4', 'FC6', 'FT8', 'C2', ' C4', 'C6', 'T8', 'CPz', 'CP2', 'CP4', 'CP6', 'TP8', 'TP10', 'P2', 'P4', 'P6', 'P8', 'POz', 'PO4', 'PO8', 'O2'. There are 9 EEG features, namely, the delta wave energy feature (denoted as DeltaPower), theta wave energy feature (denoted as ThetaPower), alpha wave energy feature (denoted as AlphaPower), alpha wave peak feature (denoted as AlphaPowerPeak), beta wave energy feature (denoted as BetaPower), non-periodic offset feature (denoted as aAperiodic), non-periodic stretching transformation feature (denoted as bAperiodic), alpha wave frequency band energy feature (denoted as AlphaPeriodicPower) and alpha wave frequency band peak feature (denoted as AlphaPeriodicPowerPeak) in the aforementioned embodiments. Therefore, there are 61*9=549 features that need to be sorted.

[0180] For sensitive brain areas, the number of times a specific signal collection position appears in the first 50 of each sorting result from rank 1 to rank 5 (a total of 549 features) (or the first 9% of all features can be selected, which is the first 9% of 549 in this embodiment) is counted. The more times a specific signal collection position appears, the more important the signal collection position is, and it is relatively sensitive to fatigue. The EEG signal collected at this signal collection position helps to improve the accuracy of fatigue detection. For example, "Signal Collection Position 1" appears 10 times and "Signal Collection Position 2" appears 6 times in the top 50 of rank 1, then the importance of "Signal Collection Position 1" is greater than that of "Signal Collection Position 2", and so on. The signal collection positions with advantages in rank 1, rank 2, rank 3, rank 4 and rank 5 can be obtained respectively, and then the sensitive brain areas can be obtained. Preferably, the signal collection positions with advantages in rank 1, rank 2, rank 3, rank 4 and rank 5 are taken as a union, and the optimal signal collection position under multiple sorting methods can be obtained, and then the sensitive brain areas that are more conducive to accurate fatigue detection can be determined.

[0181] For the dominant EEG features used for fatigue detection, the number of occurrences of the above 9 types of EEG features can also be counted in the top 50 in each sorting result in rank1, rank2, rank3, rank4 and rank5 (or the top 9% of all features can be selected, which is the top 9% of 549 in this embodiment). The more times a particular type of EEG feature appears, the more important the feature is and the more sensitive it is to fatigue. For example, if the δ wave energy feature appears 10 times and the θ wave energy feature appears 6 times in the top 50 of rank1, then the importance of the δ wave energy feature is greater than that of the θ wave energy feature, and so on, the dominant features of rank1, rank2, rank3, rank4 and rank5 can be realized respectively. Combining the five sorting results, we can get the following: Fig.17 Statistical diagram of . Fig.17 It can be seen that the α wave band energy feature and α wave band peak feature calculated by the present invention are more sensitive to fatigue than the α wave energy feature and α wave peak feature traditionally used in the prior art. Preferably, the EEG features with their respective advantages of rank 1, rank 2, rank 3, rank 4 and rank 5 are combined to obtain the optimal EEG features under multiple ranking methods, and then the type of EEG features that are more helpful for accurate fatigue detection can be determined.

[0182] When constructing a fatigue detection model, in this embodiment, first, based on the maximum number of channels allowed for fatigue detection and the sorting results obtained by the aforementioned feature selection algorithm (which can be the sorting results obtained by one feature selection algorithm or the sorting results obtained by a combination of multiple feature selection algorithms), the signal collection position that will be initially input into the fatigue detection model is selected; or a combination of multiple signal collection positions can be selected based on the specific usage environment.

[0183] Subsequently, full-feature modeling is performed, that is, the types of EEG features initially input to the fatigue detection model are the above 9 types. For the data sets used for training and testing models, the data sets can be divided into training sets, validation sets, and test sets, where the data in the test set are not included in the training set and validation set. The training set is used to train the model. The test set is used to evaluate the generalization ability of the model. After the model is trained on the training set, the test set can be used to test its performance, such as calculating indicators such as accuracy and recall, or obtaining the corresponding confusion matrix. The validation set is used for model selection and hyperparameter adjustment, which helps avoid over-optimization of the model on the training set and ensures that the model performs well on unseen data.

[0184] After the fatigue detection model is built, you can compare the accuracy of the test set to see if it meets the accuracy requirements for the actual application of the model (i.e., the accuracy requirements). If not, you can return to increase or decrease the number of signal acquisition locations used by the fatigue detection model until the accuracy requirements are met. After determining the number and type of signal acquisition locations, you can reduce the EEG features used by the fatigue detection model based on the sorting results obtained above, giving priority to reducing EEG features with low contributions, and at the same time check whether the accuracy meets the accuracy requirements. If so, you can further reduce the features. Using the above method, a fatigue detection model can be constructed with a relatively small number of signal acquisition locations, relatively few selected EEG features, low computational complexity, and that meets the accuracy requirements.

[0185] Specific experiments are introduced below to demonstrate the effectiveness of the methods of the above embodiments of the present invention, as well as the accuracy of the sensitive brain areas determined by the methods of the above embodiments and the EEG features used for fatigue detection.

[0186] In the verification experiment, a public dataset was used. The dataset comes from the paper "A resting-state EEG dataset for sleep deprivation" published in the journal ScientificData in 2024. The specific introduction is as follows:

[0187] Regarding the participants, the experiment involved 71 participants with an average age of 20 years (standard deviation 1.44 years) and an age range of 17-23 years. The gender was 34 females and 37 males. All participants reported no mental illness, anxiety or depression symptoms, recent cold symptoms, or sleep problems such as insomnia, sudden awakenings, or breathing difficulties.

[0188] Regarding the experimental design, the experiment was conducted at the Sleep and Neuroimaging Center of Southwest University from March 2019 to October 2021. The experiment adopted a within-subject design, that is, each participant experienced two conditions: normal sleep (NS) and sleep deprivation (SD). The two conditions were kept consistent as much as possible during the time period of the day (morning or afternoon), and the time difference between NS and SD for most participants was less than 1.5 hours. The interval between the two conditions was at least 7 days and at most one month to eliminate the sequential effect. The duration of sleep deprivation was 24 to 30 hours.

[0189] The experimental process was that participants were tested under NS and SD conditions in the laboratory. Before the EEG test, participants monitored their sleep quality through a sleep diary or actigraphy to ensure the quality of sleep before the experiment. Participants took the Psychomotor Vigilance Task (PVT) test, as well as scales to assess mood and sleepiness, including Positive and Negative Affect Scale (PANAS), Automatic Thoughts Questionnaire (ATQ), State Anxiety Inventory (SAI), Stanford Sleepiness Scale (SSS), Karolinska Sleepiness Scale (KSS) and a simplified sleep diary. When collecting EEG data, participants were asked to sit still with their eyes open for 5 minutes and then close their eyes for 5 minutes (some participants). 61 Ag / AgCl active electrodes were used, arranged according to the extended 10-20 international electrode placement system, FCz was used as an online reference electrode, the sampling rate was 500Hz, and the electrode impedance was kept below 5kΩ.

[0190] Sleep deprivation (SD) conditions were as follows: participants arrived at the laboratory at 21:00 the night before the experiment. During the entire deprivation period, participants were continuously monitored by two experimenters to ensure that they remained awake until the end of the experiment the next day. Participants were prohibited from consuming beverages or food containing caffeine, tea, or alcohol, and were not allowed to lie down, sleep, or engage in strenuous exercise. There was no recovery sleep period after SD, and participants underwent the experiment immediately.

[0191] The data were recorded as follows: The acquired dataset contained EEG data of 71 participants with eyes open and 38 participants with eyes closed, with the recording time standardized to 5 minutes. The uploaded data was not preprocessed.

[0192] The data preprocessing and feature extraction method is as follows: the signal acquisition position is the above 61 positions, the collected EEG signals are preprocessed according to the data segmentation, de-averaging, filtering, and windowing in the above embodiments of the present invention, the above 9 EEG features are calculated, the above five feature selection algorithms are used for feature extraction and sorting, a fatigue detection model is constructed, and machine learning training is performed. Finally, 2100 awake EEG data and 2163 fatigue EEG data are obtained, and the paired t-test Bonferroni correction method is used for significance correction, and the statistical p value and t value of the above nine EEG features are obtained. Among them, the smaller the p value, the better the distinction between awakeness and fatigue of this type of EEG feature, that is, in the brain topography corresponding to the statistical p value, the larger the coverage area of ​​the blue area, the better, and the wider the range of the t value, the better, which can indicate that the effective brain area of ​​this type of EEG feature is more extensive.

[0193] The results obtained are as follows: Fig. 8A Brain topography showing the statistical p-value corresponding to the delta wave energy feature, Figure 8B Brain topography showing the statistical t-values ​​corresponding to the delta wave energy features. Fig.9A Brain topography showing the statistical p-value corresponding to theta wave energy characteristics, Fig. 9B A brain topography map showing the statistical t-values ​​corresponding to theta wave power features. Fig. 10A Brain topography showing the statistical p-value corresponding to the alpha wave energy feature, Fig. 10B Brain topography showing the statistical t-values ​​corresponding to the alpha wave power characteristics. Fig.11A Brain topography showing the statistical p-value corresponding to the alpha wave peak energy feature, Fig. 11B A brain topography map showing the statistical t-value corresponding to the alpha wave peak energy feature. Fig. 12A Brain topography showing the statistical p-value corresponding to the beta wave energy feature, Fig. 12B Brain topography showing the statistical t-values ​​corresponding to the beta wave energy features. Fig.13A Brain topography showing the statistical p-values ​​corresponding to the non-periodic telescoping transformation features, Fig. 13B Brain topography showing the statistical t-values ​​corresponding to the non-periodic stretching transformation features. Fig.14A Brain topography showing the statistical p-values ​​corresponding to the non-periodic excursion features, Fig. 14B Brain topography showing the statistical t-values ​​corresponding to the non-periodic excursion features. Fig.15A Brain topography showing the statistical p-value corresponding to the alpha wave band energy characteristics, Fig. 15BA brain topography map showing the statistical t-value corresponding to the energy characteristics of the α wave band. Fig.16A Brain topography showing the statistical p-value corresponding to the peak characteristics of the alpha wave band, Fig. 16B The brain topography map showing the statistical t value corresponding to the α wave band peak feature. For example, it can be seen that the dominant brain area corresponding to the α wave band energy feature and the α wave band peak feature covers a wide area, and the absolute value of the t value can reach greater than 20, and the t value range is wider.

[0194] Table 1: Classification accuracy statistics of 61 channels using 5 traditional features and 9 features of the present invention

[0195]

[0196] In order to more objectively compare the effectiveness of the features, 34 machine learning models were also used to process the feature data, see Table 1. It can be seen from Table 1 that the classification accuracy after using the features proposed by the present invention has increased from the original 90.07% to 90.25%, proving the effectiveness of the present invention. In Table 1, the traditional five features refer to 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, and the nine features of the present invention refer to the δ wave energy feature, the θ wave energy feature, the α wave energy feature, the α wave peak feature, the β wave energy feature, the non-periodic offset feature, the non-periodic stretching transformation feature, the α wave frequency band energy feature, and the α wave frequency band peak feature.

[0197] Furthermore, the classification accuracy results show that the dominant brain regions proposed by the present invention have higher classification accuracy than the traditional recognition accuracy of randomly selecting several brain regions. Figure 18-Figure 23 The dominant brain area statistics are shown. Figure 18-Figure 23 More specifically, it shows the degree of difference between the same type of awake EEG characteristics and fatigue EEG characteristics in each area. The redder the color, the greater the difference, which means that the signal collected in this area is more sensitive to fatigue and can be used as a sensitive brain area.

[0198] in, Fig.18 The figure shows the brain topography of sensitive brain areas obtained by using the chi-square test algorithm, which is specifically drawn according to the number of occurrences of different signal acquisition positions in the sorting results obtained by using the chi-square test algorithm; Fig.19 The brain topography of sensitive brain areas obtained by using the minimum redundancy maximum correlation (MRMR) algorithm is shown, specifically, it is drawn according to the number of times different signal acquisition positions appear in the sorting results obtained by using the minimum redundancy maximum correlation (MRMR) algorithm; Fig. 20The brain topography of the sensitive brain area obtained by using the Kruskal-Wallis test algorithm is shown, which is specifically drawn according to the number of occurrences of different signal acquisition positions in the sorting results obtained by using the Kruskal-Wallis test algorithm; Fig.21 The brain topography of sensitive brain areas obtained by using the ReliefF algorithm is shown, which is specifically drawn according to the number of occurrences of different signal collection positions in the sorting results obtained by using the ReliefF algorithm; Fig. 22 The figure shows the brain topography of sensitive brain areas obtained by using the recursive feature ranking algorithm based on support vector machine (SVM-REF), which is specifically drawn according to the number of occurrences of different signal acquisition positions in the ranking results obtained by using the recursive feature ranking algorithm based on support vector machine (SVM-REF); Fig.23 The brain topography map of sensitive brain areas obtained by combining five feature selection algorithms in the embodiment of the present invention is shown. Specifically, it is drawn according to the total number of times different signal collection positions appear in the sorting results obtained by the above five feature selection algorithms, that is, it is drawn according to the union of the first five feature selection algorithms. Among them, the sensitive brain area is the area with reddish color. Reference Figure 18-Figure 23 From the dominant brain area statistics shown, it can be seen that F3, F4 and O1, O2, Oz are the dominant brain areas (i.e., sensitive brain areas), that is, the sensitive brain areas finally determined are the five signal collection locations of F3, F4 and O1, O2, Oz.

[0199] Referring to Table 2, the average accuracy of the 5 electrodes in the dominant brain area (F3, F4, O1, O2, Oz) is 4.2% higher than that of the random 5 electrodes (the random 5 electrodes excluding F3, F4, O1, O2, Oz). The sensitive brain area selected by the above method of the present invention is effective for fatigue detection.

[0200] Table 2: Classification accuracy statistics of 5 electrodes in the dominant brain area (F3, F4, O1, O2, Oz) and 5 random electrodes

[0201]

[0202] Second, reference Figure 4 The embodiment of the present invention discloses a device 1 for determining sensitive brain areas and EEG features for fatigue detection, wherein the sensitive brain areas and EEG features for fatigue detection are determined by the method for determining sensitive brain areas and EEG features for fatigue detection in any of the aforementioned embodiments. The device 1 for determining sensitive brain areas and EEG features for fatigue detection comprises a signal acquisition module 11, a feature calculation module 12, and a sensitive brain area and EEG feature determination module 13.

[0203] The signal acquisition module 11 is connected to the feature calculation module 12 , and the support image acquisition module 12 is connected to the sensitive brain area and EEG feature determination module 13 .

[0204] Continue to refer Figure 4 The signal acquisition module 11 is used to select multiple signal acquisition positions on the head and collect awake EEG signals and fatigue EEG signals corresponding to each signal acquisition position. The awake EEG signals are EEG signals in the awake state, and the fatigue EEG signals are EEG signals in the fatigue state.

[0205] The feature calculation module 12 is used to determine a plurality of awake EEG features corresponding to the awake EEG signals and a plurality of fatigue EEG features corresponding to the fatigue EEG signals according to the awake EEG signals and fatigue EEG signals collected at each signal collection position.

[0206] The sensitive brain region and EEG feature determination module 13 is used to determine the degree of difference between the same type of awake EEG features and fatigue EEG features at each signal acquisition position, and sort the signal acquisition positions and the corresponding EEG features according to the degree of difference to obtain a sorting result, and select the final signal acquisition position and the final EEG feature as the sensitive brain region and the EEG feature for fatigue judgment from the signal acquisition positions and EEG features within the preset high difference degree range according to the sorting result. Among them, the preset high difference degree range is a range with a relatively high difference degree in the sorting.

[0207] By adopting the above technical scheme, the device 1 for determining sensitive brain areas and EEG features for fatigue detection of the present invention can compare the degree of difference between the same type of awake EEG features and fatigue EEG features collected at different signal collection positions, and can also compare the sensitivity of different types of EEG features to fatigue, and then select the signal collection positions and EEG features within a preset high difference range by sorting, and then select sensitive brain areas and EEG features for fatigue judgment. It can accurately select sensitive brain areas and EEG features suitable for fatigue detection, and can also make the selected sensitive brain areas relatively small, and the selected EEG features relatively easier to calculate.

[0208] Furthermore, the signal acquisition module 11 can also be used to execute the aforementioned steps S11, S12 and related steps. The feature calculation module 12 can also be used to execute the steps of determining the non-periodic offset characteristics, non-periodic stretching transformation characteristics, α wave band energy characteristics and α wave band peak characteristics of the EEG signal and related steps, such as steps S01, S02, S03, S04, S05, S06 and related steps. The sensitive brain area and EEG feature determination module 13 can also be used to execute the aforementioned steps S31, S32, S33, S34, S35, S36, S37 and related steps.

[0209] In a third aspect, an embodiment of the present invention discloses a fatigue detection method, which performs fatigue detection using the sensitive brain areas and EEG features determined by the method for determining sensitive brain areas and EEG features for fatigue detection in any of the aforementioned embodiments.

[0210] By adopting the above technical solution, the fatigue detection method of the present invention can achieve accurate fatigue detection through as few sensitive brain areas and EEG features as possible, and can reduce the amount of calculation. More specifically, the fatigue detection method can reduce the number of channels for collecting EEG signals, simplify the data collection method, simplify the calculation of EEG features, reduce the calculation complexity, and shorten the calculation time, thereby improving the convenience of fatigue detection.

[0211] Furthermore, the determined sensitive brain areas include the F3, F4, O1, O2, and Oz electrode positions of the 10-20 international standard lead system, and the determined EEG characteristics include delta wave energy characteristics, theta wave energy characteristics, alpha wave energy characteristics, alpha wave peak characteristics, beta wave energy characteristics, non-periodic offset characteristics, non-periodic stretching and transformation characteristics, alpha wave frequency band energy characteristics, and alpha wave frequency band peak characteristics.

[0212] Among them, by using different numbers or types of feature selection algorithms to obtain sorting results, and then selecting sensitive brain areas and EEG features for fatigue judgment, sensitive brain areas and EEG features different from the above-mentioned five electrode positions and nine EEG features may be selected. In addition, for different people, different data sets, and different application scenarios, other sensitive brain areas and EEG features for fatigue detection may be selected through the method for determining sensitive brain areas and EEG features for fatigue detection in any of the aforementioned embodiments. In the present invention, the above-mentioned sensitive brain areas and EEG features (i.e., the above-mentioned five electrode positions and nine EEG features) determined by the method for determining sensitive brain areas and EEG features for fatigue detection in any of the aforementioned embodiments meet the needs of fatigue detection in most cases and are universal. When fatigue detection is actually performed, it is preferred to retain the above-mentioned determined sensitive brain areas and EEG features.

[0213] Fourthly, reference Figure 5 The embodiment of the present invention further discloses an electronic device 2, including a memory 21 and a processor 22, wherein the memory 21 stores at least one instruction, and when the at least one instruction is executed by the processor 22, it implements the method for determining the sensitive brain area and EEG characteristics for fatigue detection in any embodiment, and / or implements the fatigue detection method in any of the aforementioned embodiments. Among them, the memory 21 may include, for example, a system memory, a fixed non-volatile storage medium, etc. The system memory, for example, stores an operating system, an application, a boot loader, and other programs.

[0214] In this embodiment, the electronic device 2 can compare the degree of difference between the same type of awake EEG features and fatigue EEG features collected at different signal collection positions, and can also compare the sensitivity of different types of EEG features to fatigue, and then select the signal collection positions and EEG features within the preset high difference range by sorting, and then select sensitive brain areas and EEG features for fatigue judgment, which can accurately select sensitive brain areas and EEG features suitable for fatigue detection, and can also make the selected sensitive brain areas relatively small, and the selected EEG features are relatively easier to calculate. In addition, when performing fatigue detection, accurate fatigue detection can be achieved with as few sensitive brain areas and EEG features as possible, and the amount of calculation can be reduced.

[0215] In a fifth 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, it implements the method for determining sensitive brain areas and EEG characteristics for fatigue detection in any of the aforementioned embodiments, and / or implements the fatigue detection method in any of the aforementioned embodiments.

[0216] By adopting the above technical solution, the computer-readable storage medium can compare the degree of difference between the same type of awake EEG features and fatigue EEG features collected at different signal collection positions, and can also compare the sensitivity of different types of EEG features to fatigue, and then select the signal collection positions and EEG features within the preset high difference range by sorting, and then select sensitive brain areas and EEG features for fatigue judgment, which can accurately select sensitive brain areas and EEG features suitable for fatigue detection, and can also make the selected sensitive brain areas relatively small, and the selected EEG features are relatively easier to calculate. In addition, when performing fatigue detection, accurate fatigue detection can be achieved with as few sensitive brain areas and EEG features as possible, and the amount of calculation can be reduced.

[0217] 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.

[0218] 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.

[0219] 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 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in 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.

[0220] 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 method for determining sensitive brain areas and EEG features for fatigue detection, characterized in that: The sensitive brain area is the area of ​​the head where the EEG characteristics are collected, and the determination method includes: Selecting multiple signal collection positions on the head and collecting awake EEG signals and fatigue EEG signals corresponding to each signal collection position; wherein the awake EEG signals are EEG signals in an awake state, and the fatigue EEG signals are EEG signals in a fatigue state; Determine, according to the awake EEG signals and fatigue EEG signals collected at each signal collection position, a plurality of awake EEG features corresponding to the awake EEG signals and a plurality of fatigue EEG features corresponding to the fatigue EEG signals; Determine the degree of difference between the awake EEG features and the fatigue EEG features of the same type 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, and select the final signal collection position and the final EEG feature as the sensitive brain area and the EEG feature for fatigue judgment from the signal collection positions and EEG features within a preset high difference degree range according to the sorting result; wherein the preset high difference degree range is a range with a relatively high difference degree in the sorting.

2. The method for determining sensitive brain areas and EEG features for fatigue detection according to claim 1, characterized in that: The types of the awake EEG feature and the fatigue EEG feature include aperiodic offset feature and aperiodic stretching and transforming feature. The aperiodic offset feature and the aperiodic stretching and transforming feature of the EEG signal are 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 offset parameter a in the formula 1 is determined to be the non-periodic offset feature, and the scaling parameter b in the formula 1 is determined to be the non-periodic scaling feature.

3. The method for determining sensitive brain areas and EEG features for fatigue detection according to claim 2, characterized in that: The types of awake EEG features and fatigue EEG features also include α wave band energy features and α wave band peak features. The α wave band energy features and α wave band peak features of the EEG signal are determined by: 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 characteristic and the α wave frequency band peak characteristic are determined according to the α wave spectrum period energy component.

4. The method for determining sensitive brain areas and EEG features for fatigue detection according to claim 1, characterized in that: The step of collecting the corresponding awake EEG signal and fatigue EEG signal at each of the multiple signal collection positions on the head includes: At each of the multiple signal collection positions on the head, a corresponding original awake EEG signal and an original fatigue EEG signal are respectively collected; The original awake EEG signal and the original fatigue EEG signal are preprocessed to obtain the awake EEG signal and the fatigue EEG signal; wherein the preprocessing method includes at least one of the following: data segmentation, de-averaging, filtering, and windowing.

5. The method for determining sensitive brain areas and EEG features for fatigue detection according to claim 1, characterized in that: The degree of difference between the awake EEG features and the fatigue EEG features of the same type is determined by a feature selection algorithm, and the final signal collection position and the final EEG feature are selected from the signal collection positions and EEG features within a preset high difference degree range according to the sorting result, specifically including: Determining the number of occurrences of different signal collection positions and different EEG features within the preset high difference degree range; The signal collection positions and EEG features with relatively large number of occurrences are selected as the position and feature candidate data sets corresponding to the feature selection algorithm, and the final signal collection position and the final EEG feature are selected from the position and feature candidate data sets.

6. The method for determining sensitive brain areas and EEG features for fatigue detection according to claim 5, characterized in that: There are multiple feature selection algorithms, and the final signal collection position and the final EEG feature are selected from the signal collection positions and EEG features within a preset high difference range according to the sorting result, and further includes: Obtaining the position and feature selection dataset corresponding to each feature selection algorithm; In the union of the plurality of position and feature candidate data sets, the signal acquisition position and the EEG feature with relatively more occurrence times are selected as the final signal acquisition position and the final EEG feature.

7. The method for determining sensitive brain areas and EEG features for fatigue detection according to claim 6, characterized in that: The multiple feature selection algorithms include a feature selection algorithm based on machine learning and a feature selection algorithm based on statistics.

8. The method for determining sensitive brain areas and EEG features for fatigue detection according to claim 7, characterized in that: The multiple feature selection algorithms include minimum redundancy maximum relevance (MRMR) algorithm, chi-square test algorithm, ReliefF algorithm, Kruskal-Wallis test algorithm, and support vector machine-based recursive feature sorting algorithm SVM-REF.

9. The method for determining sensitive brain areas and EEG features for fatigue detection according to claim 1, characterized in that: The selecting the final signal collection position and the final EEG feature from the signal collection positions and EEG features within a preset high difference range according to the sorting result specifically includes: Acquire a fatigue detection model, wherein the fatigue detection model performs fatigue detection based on signal acquisition position and EEG characteristics; Selecting a pending signal collection position and a pending EEG feature from within the preset high difference degree range, inputting the pending signal collection position and the pending EEG feature into the fatigue detection model and performing fatigue detection, and determining the accuracy of fatigue judgment by the fatigue detection model; The number and / or type of the pending signal collection positions and the pending EEG features are changed, and the changed pending signal collection positions and pending EEG features are input into the fatigue detection model to determine the accuracy of fatigue judgment, until the minimum number of pending signal collection positions and pending EEG features whose determined accuracy meets the preset high accuracy condition are obtained, and the pending signal collection positions and pending EEG features corresponding to the last change are determined as the final signal collection positions and the final EEG features.

10. A device for determining sensitive brain areas and EEG characteristics for fatigue detection, characterized in that: The sensitive brain area is the area of ​​the head where the EEG characteristics are collected, and the determining device includes: A signal acquisition module, used to select multiple signal acquisition positions on the head and acquire awake EEG signals and fatigue EEG signals corresponding to each signal acquisition position; wherein the awake EEG signals are EEG signals in an awake state, and the fatigue EEG signals are EEG signals in a fatigue state; A feature calculation module, used for determining a plurality of awake EEG features corresponding to the awake EEG signals and a plurality of fatigue EEG features corresponding to the fatigue EEG signals according to the awake EEG signals and fatigue EEG signals collected at each signal collection position; The sensitive brain area and EEG feature determination module is used to determine the degree of difference between the awake EEG features and the fatigue EEG features of the same type at each signal collection position, and sort the signal collection positions and the corresponding EEG features according to the size of the difference degree to obtain a sorting result, and select the final signal collection position and the final EEG feature as the sensitive brain area and the EEG feature for fatigue judgment from the signal collection positions and EEG features within a preset high difference degree range according to the sorting result; wherein the preset high difference degree range is a range with a relatively high difference degree in the sorting.

11. A fatigue detection method, characterized in that: Fatigue detection is performed using the sensitive brain area and EEG features determined by the method for determining the sensitive brain area and EEG features for fatigue detection as described in any one of claims 1 to 9.

12. The fatigue detection method according to claim 11, characterized in that: The determined sensitive brain areas include the F3, F4, O1, O2, and Oz electrode positions of the 10-20 international standard lead system, and the determined EEG characteristics include delta wave energy characteristics, theta wave energy characteristics, alpha wave energy characteristics, alpha wave peak characteristics, beta wave energy characteristics, non-periodic offset characteristics, non-periodic stretching and transformation characteristics, alpha wave frequency band energy characteristics, and alpha wave frequency band peak characteristics.

13. 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, it implements the method for determining sensitive brain areas and EEG characteristics for fatigue detection as described in any one of claims 1-9, and / or implements the fatigue detection method as described in any one of claims 11-12.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, which, when executed, implements the method for determining sensitive brain areas and EEG characteristics for fatigue detection as described in any one of claims 1-9, and / or implements the fatigue detection method as described in any one of claims 11-12.

Images