Electroencephalogram-electrooculogram-based driving fatigue recognition method based on uncertainty evidence fusion, computer device and storage medium

Abstract

The application discloses a driving fatigue recognition method based on electroencephalogram-eyeground electrogram uncertainty evidence fusion, a computer device and a storage medium, relates to the technical field of intelligent driving assistance systems, and comprises the following steps: collecting electroencephalogram signals and eyeground electrogram signals of a driver in a sober state, constructing a personalized physiological baseline containing multi-dimensional characteristic statistical parameters, and setting a dynamic fatigue determination threshold based on the personalized physiological baseline; in the driving process, real-time electroencephalogram signals, real-time eyeground electrogram signals and context information reflecting the current driving environment and the vehicle operating state are synchronously acquired; by using the sober period data to establish the dynamic determination threshold, precise adaptation to the physiological state of the driver is realized, the weight of the disturbed mode is dynamically inhibited by introducing the context information, and the mode uncertainty is calibrated through the vehicle behavior index, so that true fatigue and false abnormalities can be effectively distinguished, the relationship among the support degree, the uncertainty and the conflict degree is explicitly modeled by using the Dempster-Shafer fusion rule, reasonable confidence judgment can still be output in a high conflict scene, and false positives or false negatives are avoided.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving assistance system technology, and in particular to a driving fatigue recognition method, computer equipment and storage medium based on the fusion of uncertain evidence from electroencephalography and electrooculography. Background Technology

[0002] Intelligent driving assistance system technology refers to a comprehensive technology system that uses sensors, artificial intelligence algorithms and vehicle control technology to perceive the vehicle's surrounding environment in real time and provide warnings or proactive intervention to the driver when necessary, so as to improve driving safety and driving comfort.

[0003] Traditional methods often use general thresholds or fixed models, which are difficult to adapt to the differences in physiological baselines of different drivers in a conscious state, resulting in high false alarm rates or insufficient sensitivity. In real driving scenarios, factors such as changes in lighting, road bumps, and violent vehicle operations can seriously contaminate electrooculography (EOG) or electroencephalography (EEG) signals, and existing systems lack dynamic evaluation and suppression mechanisms for signal quality. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method, computer device, and storage medium for driver fatigue recognition based on the fusion of uncertain evidence from EEG and EEG. This solves the problem that traditional methods often use general thresholds or fixed models, which are difficult to adapt to the differences in the physiological baseline of different drivers in a conscious state, resulting in high false alarm rates or insufficient sensitivity. Furthermore, in real driving scenarios, factors such as changes in lighting, road bumps, and violent vehicle operations can seriously contaminate EEG or EEG signals, and existing systems lack dynamic evaluation and suppression mechanisms for signal quality.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for driver fatigue recognition based on EEG-EOG uncertainty evidence fusion, comprising: Collect EEG and EEG signals of drivers in a conscious state, construct a personalized physiological baseline containing multidimensional feature statistical parameters, and set a dynamic fatigue judgment threshold based on the personalized physiological baseline. During driving, real-time EEG signals, real-time EOG signals, and contextual information reflecting the current driving environment and vehicle operating status are acquired simultaneously. The prior reliability weights of EEG and EEG modalities are determined based on contextual information. Real-time EEG and EEG signals are preprocessed and depersonalized to obtain EEG feature tensors and EEG feature vectors adapted to individual characteristics. The EEG feature tensor and EEG feature vector are respectively input into the corresponding pre-trained branch model to generate the original EEG evidence vector and the original EEG evidence vector. The original EEG evidence vector and the original EEG evidence vector are modulated by combining the prior reliability weight to obtain the modulated EEG evidence vector and the modulated EEG evidence vector. The initial modal uncertainty is calculated based on the modulated EEG evidence vector and the modulated EEG evidence vector, and the initial modal uncertainty is calibrated online through a lightweight calibration network using the vehicle behavior stability index, outputting the calibrated EEG uncertainty and the calibrated EEG uncertainty. Based on the modulated EEG evidence vector, modulated EEG evidence vector, and calibrated EEG and EEG uncertainties, corresponding belief quality is generated. The Dempster-Shafer evidence theory is used to fuse the belief quality, output the fused fatigue state, and perform graded early warning based on the fused fatigue state.

[0007] As a preferred embodiment of the driver fatigue recognition method based on EEG-EOG uncertainty evidence fusion described in this invention, the construction of a personalized physiological baseline containing multidimensional feature statistical parameters includes the following steps: EEG and EOG signals were continuously collected while the driver was awake and at rest, with a sampling time of no less than five minutes. The EEG signal was segmented and windowed, with each window lasting two seconds and having a 50% overlap. The differential entropy and power spectral density were calculated for each segment. The energy proportion of each frequency band, along with the sample entropy, forms the EEG feature vector; Waveform detection is performed on the electrooculogram (EOG) signal to extract the duration of a single blink, the interval between adjacent blinks, and the slow-phase eye movement velocity, which constitute the EOG feature vector. Calculate the mean of all EEG feature vectors by dimension. with standard deviation Calculate the mean of all electrooculography feature vectors by dimension. with standard deviation These combinations form a personalized physiological baseline; in, The first characteristic of EEG The mean of the dimensionality in a conscious state. Indicates the corresponding standard deviation. The first characteristic of electrooculography The mean of the dimensionality in a conscious state. This represents the corresponding standard deviation.

[0008] As a preferred embodiment of the driver fatigue recognition method based on EEG-EOG uncertainty evidence fusion described in this invention, the step of setting a dynamic fatigue judgment threshold based on a personalized physiological baseline includes the following steps: The first of the EEG characteristics Dimensionally set upper and lower threshold values ​​for the electrooculogram feature. Define upper and lower bound thresholds; When the real-time extracted EEG or EEG features exceed the upper and lower bound thresholds of the corresponding dimension, the dimension is marked as a fatigue-sensitive dimension. The upper and lower bound thresholds are set using a uniform scaling factor. ,satisfy ; EEG characteristics The upper bound threshold of dimension is The lower bound threshold is ; Electroocular characteristics The upper bound threshold of dimension is The lower bound threshold is ; in, This is a preset constant used to control the tolerance bandwidth of an individual baseline.

[0009] As a preferred embodiment of the driver fatigue recognition method based on EEG-EOG uncertainty evidence fusion described in this invention, the step of determining the prior reliability weights of EEG modalities and EOG modalities based on contextual information includes the following steps: The vehicle obtains current light intensity, vehicle speed, longitudinal acceleration, and lateral acceleration from the vehicle's CAN bus, and calculates the road bump index from the suspension vibration signal; If the light intensity is below the dark light threshold, the prior reliability weight of the electrooculogram modality is set to an attenuation factor of less than one; otherwise, it is set to one. If the road surface bump index is higher than the bump threshold, the weight of the prior reliability of the EEG modality is set to a decay factor of less than one; otherwise, it is set to one. The prior reliability weight is used to modulate the original evidence vector subsequently, reflecting the degree of influence of the external environment on the quality of physiological signals.

[0010] As a preferred embodiment of the driver fatigue recognition method, computer device, and storage medium based on EEG-EOG uncertainty evidence fusion described in this invention, the depersonalization process includes the following steps: Perform a standardization transformation on each dimension of the real-time EEG feature vector, that is, subtract the corresponding baseline mean from the feature value of that dimension and divide by the corresponding baseline standard deviation to obtain the depersonalized EEG feature tensor; Perform the same standardized transformation on each dimension of the real-time electrooculogram feature vector to obtain a depersonalized electrooculogram feature vector. The standardization transformation maps the feature representations of different drivers to the same relative offset space, eliminating absolute magnitude differences between individuals.

[0011] As a preferred embodiment of the driver fatigue recognition method based on EEG-EOG uncertainty evidence fusion described in this invention, the step of inputting the EEG feature tensor and EOG feature vector into the corresponding pre-trained branch model to generate the original EEG evidence vector and the original EOG evidence vector, and modulating the original EEG evidence vector and the original EOG evidence vector with prior reliability weights to obtain the modulated EEG evidence vector and the modulated EOG evidence vector, includes the following steps: The depersonalized EEG feature tensor is input into a pre-trained convolutional neural network branch, which outputs the raw EEG evidence vector. ; The depersonalized electrooculogram (EOG) feature vector is input into a pre-trained fully connected neural network branch, which outputs the original EOG evidence vector. ; The original EEG evidence vector and the prior reliability weights of EEG modalities are combined. Multiplying them yields the modulated EEG evidence vector. ; The original electrooculogram evidence vector and the electrooculogram modal prior reliability weights are combined. Multiplying them together yields the modulated electrooculography evidence vector. ; in, This indicates the total number of fatigue state categories. This indicates the strength of the original evidence from the EEG branches regarding type I fatigue. Indicates the effect of the electrooculography branch on the first The strength of the original evidence for fatigue-like states. and To take values ​​in Real weights for intervals.

[0012] As a preferred embodiment of the driving fatigue recognition method based on EEG-EOG uncertainty evidence fusion described in this invention, the step of generating corresponding belief quality based on modulated EEG evidence vector, modulated EOG evidence vector, calibrated EEG uncertainty, and calibrated EOG uncertainty, fusing the belief quality using Dempster-Shafer evidence theory, and outputting the fused fatigue state includes the following steps: Modulated EEG Evidence Vector Each element is summed and normalized to obtain the EEG support. ,in ; Modulated electrooculo evidence vector Each element is summed and normalized to obtain the electrooculogram support. ,in ; Calculating EEG uncertainty Electrooculometry uncertainty ; Calculate the evidence conflict coefficient between the two modalities ; like Then, after calculation and fusion, the first... Support for fatigue-like states With fusion uncertainty The formula is as follows: ; ; In the formula, This indicates the degree of support from EEG modalities for the first type of fatigue state. This indicates the degree of support of the electrooculogram (EOG) mode for the type I fatigue state. This represents the prediction uncertainty of EEG modalities. This represents the prediction uncertainty of the electrooculogram (EOG) mode. This indicates the degree of evidence conflict between the two modalities; the output after fusion is the fatigue state. .

[0013] As a preferred embodiment of the driver fatigue recognition method based on EEG-EOG uncertainty evidence fusion described in this invention, the step of performing graded early warning based on the fused fatigue state includes the following steps: Set the first uncertainty threshold With the second uncertainty threshold ,satisfy ; When the fatigue state after fusion is severe fatigue and the fusion uncertainty is... When this occurs, strong intervention measures are triggered, including audible alarms, seat vibration, and red flashing on the dashboard; When the fatigue state after fusion is mild fatigue, or the fatigue state after fusion is severe fatigue but the fusion uncertainty is high. At this time, only weak prompts are triggered, including voice reminders and yellow prompts on the dashboard; When fusion uncertainty At that time, no warning was triggered.

[0014] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the driving fatigue recognition method, computer device and storage medium based on EEG-EOG uncertainty evidence fusion as described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the driving fatigue recognition method, computer device and storage medium based on EEG-EOG uncertainty evidence fusion as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By utilizing awake period data to establish a dynamic judgment threshold, it achieves precise adaptation to the driver's physiological state; it introduces contextual information to dynamically suppress the weight of the interfered modality; and it calibrates modal uncertainty through vehicle behavior indicators, effectively distinguishing between true fatigue and false anomalies. The Dempster-Shafer fusion rule explicitly models the relationship between support, uncertainty, and conflict, and can still output reasonable confidence judgments in high-conflict scenarios, avoiding false alarms or missed alarms. It implements graded early warning based on fusion uncertainty, ensuring that strong intervention is only triggered when there is high confidence and severe fatigue, taking into account both safety performance and user experience. It solves the problems of multimodal physiological signals being easily interfered with, having weak generalization ability, and having unreliable fusion results in the vehicle environment. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the driver fatigue recognition method based on EEG-EOG uncertainty evidence fusion in Example 1. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Example 1, referring to Figure 1 As one embodiment of the present invention, this embodiment provides a method for driver fatigue recognition based on EEG-EOG uncertainty evidence fusion, comprising the following steps: S1. Collect EEG and EEG signals of the driver in a conscious state, construct a personalized physiological baseline containing multidimensional feature statistical parameters, and set a dynamic fatigue judgment threshold based on the personalized physiological baseline.

[0023] Furthermore, while the driver is awake and at rest, continuous EEG and EEG signals are collected for at least five minutes. The EEG signals are segmented and windowed, with each window lasting two seconds and having a 50% overlap. The differential entropy and power spectral density are calculated for each segment. The energy proportion of the frequency band and the sample entropy are used to form the EEG feature vector; the waveform of the electrooculogram (EOG) signal is detected, and the duration of a single blink, the interval between adjacent blinks, and the slow-phase eye movement velocity are extracted to form the EOG feature vector.

[0024] Calculate the mean of all EEG feature vectors by dimension. with standard deviation Calculate the mean of all electrooculography feature vectors by dimension. with standard deviation These combinations form a personalized physiological baseline; in, The first characteristic of EEG The mean of the dimensionality in a conscious state. Indicates the corresponding standard deviation. The first characteristic of electrooculography The mean of the dimensionality in a conscious state. This represents the corresponding standard deviation.

[0025] The first of the EEG characteristics Dimensionally set upper and lower threshold values ​​for the electrooculogram feature. The system sets upper and lower bound thresholds for each dimension; when the real-time extracted EEG or EEG features exceed the corresponding upper and lower bound thresholds, that dimension is marked as a fatigue-sensitive dimension; the upper and lower bound thresholds are set using a uniform scaling factor. ,satisfy EEG characteristics The upper bound threshold of dimension is The lower bound threshold is Electroocular characteristics The upper bound threshold of dimension is The lower bound threshold is ;in, This is a preset constant used to control the tolerance bandwidth of an individual baseline.

[0026] It should be noted that by constructing a personalized physiological baseline containing multidimensional statistical features under conscious and resting conditions, and setting a dynamic fatigue judgment threshold accordingly, the interference caused by physiological differences between individuals can be effectively eliminated, making the fatigue judgment criteria truly adaptable to the normal state of each driver. At the same time, the adoption of an adaptive threshold mechanism based on the standard deviation multiple not only improves the sensitivity to early signs of fatigue, but also avoids misjudgment or missed judgment caused by fixed thresholds, thereby enhancing the system's generalization ability and robustness in different populations.

[0027] S2. During driving, real-time EEG signals, real-time EOG signals, and contextual information reflecting the current driving environment and vehicle operating status are acquired simultaneously.

[0028] Furthermore, wearable EEG acquisition devices continuously acquire multi-channel EEG signals from the driver's forehead and occipital lobe regions at a sampling rate of no less than 250 Hz; horizontal and vertical EEG signals are acquired through eye-tracking modules integrated into the steering wheel or rearview mirror, or through attached EOG electrodes; vehicle speed, engine speed, steering angle, longitudinal acceleration, lateral acceleration, brake pedal travel, and throttle opening signals are simultaneously read from the vehicle's controller local area network bus; ambient illuminance in the cockpit is obtained through a light sensor installed on the vehicle body, and the current time and geographical location are obtained through a GPS module to determine the day / night state; physiological signals and vehicle status data are aligned with a unified timestamp to form a synchronous data stream for subsequent modal reliability assessment and feature extraction.

[0029] It should be noted that by simultaneously acquiring high-sampling-rate multi-channel EEG and EEG signals and integrating multi-source contextual information from the vehicle bus, light sensor, and positioning module, the spatiotemporal alignment of physiological state and driving scenario is achieved. The multimodal synchronous perception mechanism provides a comprehensive and consistent data foundation for subsequent environmental interference identification, signal quality assessment, and uncertainty calibration, ensuring the real-time performance and reliability of the system in real and complex road environments.

[0030] S3. Determine the prior reliability weights of EEG and EEG modalities based on contextual information, and preprocess and depersonalize the real-time EEG and EEG signals to obtain EEG feature tensors and EEG feature vectors adapted to individual characteristics.

[0031] Furthermore, the system obtains the current light intensity, vehicle speed, longitudinal acceleration, and lateral acceleration from the vehicle's CAN bus, and calculates the road bump index from the suspension vibration signal. If the light intensity is lower than the dark light threshold, the prior reliability weight of the electrooculogram (EOG) mode is set to an attenuation factor of less than one; otherwise, it is set to one. If the road bump index is higher than the bump threshold, the prior reliability weight of the electroencephalogram (EEG) mode is set to an attenuation factor of less than one; otherwise, it is set to one. The prior reliability weight is used to modulate the original evidence vector in the future, reflecting the degree of influence of the external environment on the quality of physiological signals.

[0032] A standardization transformation is performed on each dimension of the real-time EEG feature vector, i.e., the feature value of that dimension is subtracted from the corresponding baseline mean and then divided by the corresponding baseline standard deviation, to obtain the depersonalized EEG feature tensor. The same standardization transformation is performed on each dimension of the real-time EEG feature vector to obtain the depersonalized EEG feature vector. The standardization transformation maps the feature representations of different drivers to the same relative offset space, eliminating the absolute magnitude differences between individuals. It should be noted that dynamically adjusting the prior reliability weights of EEG and EEG modalities based on lighting conditions and road surface conditions can proactively suppress the negative impact of modalities severely affected by environmental interference on decision-making. At the same time, by performing standardization and de-personalization processing on real-time features based on individual baselines, the original signals are mapped to a unified relative offset space, which not only preserves fatigue-related abnormal change information but also eliminates individual absolute magnitude differences, thereby improving the consistency of cross-user features and model adaptability.

[0033] S4. Input the EEG feature tensor and EEG feature vector into the corresponding pre-trained branch model to generate the original EEG evidence vector and the original EEG evidence vector. Modulate the original EEG evidence vector and the original EEG evidence vector by combining the prior reliability weight to obtain the modulated EEG evidence vector and the modulated EEG evidence vector.

[0034] Furthermore, the depersonalized EEG feature tensor is input into a pre-trained convolutional neural network branch, which outputs the raw EEG evidence vector. The depersonalized electrooculogram (EOG) feature vector is input into a pre-trained fully connected neural network branch, which outputs the original EOG evidence vector. The original EEG evidence vector is combined with the prior reliability weights of the EEG modalities. Multiplying them yields the modulated EEG evidence vector. The original electrooculogram (EOG) evidence vector is combined with the prior reliability weights of the EOG modality. Multiplying them together yields the modulated electrooculography evidence vector. ;in, This indicates the total number of fatigue state categories. Indicates the EEG branches on the first The strength of the original evidence for fatigue-like states. This indicates the strength of the original evidence from the electrooculography branch regarding type I fatigue. and To take values ​​in Real weights for intervals.

[0035] It should be noted that a specially designed pre-trained branch model is used to process EEG and EEG features separately, and the original evidence is modulated by combining the reliability weight of environmental perception. This achieves an organic combination of data-driven feature extraction and scene-aware evidence correction. The mechanism enables the system to maintain the discrimination ability of each modality while intelligently adjusting its contribution according to the current driving conditions, effectively improving the quality and credibility of the evidence before fusion.

[0036] S5. Calculate the initial modal uncertainty based on the modulated EEG evidence vector and the modulated EEG evidence vector, and use the vehicle behavior stability index to perform online calibration of the initial modal uncertainty through a lightweight calibration network, and output the calibrated EEG uncertainty and the calibrated EEG uncertainty.

[0037] Furthermore, the initial uncertainty of the EEG modality is calculated based on the modulated EEG evidence vector, which reflects the confidence level of the EEG branch model in judging the current fatigue state; the initial uncertainty of the EEG modality is calculated based on the modulated EEG evidence vector, which reflects the confidence level of the EEG branch model in judging the current fatigue state; and the vehicle speed change rate, lateral acceleration fluctuation amplitude, and steering angle change frequency are extracted from the synchronously acquired context information to generate a vehicle behavior stability index characterizing the smoothness of vehicle operation.

[0038] Initial uncertainties in EEG, EEG, and vehicle behavior stability indicators are fed into a pre-trained lightweight calibration network. This network, a simple, low-computational neural network, is used to adjust the initial uncertainties in real time to eliminate the interference of physiological signal distortion caused by violent vehicle operations on uncertainty estimation. The network outputs corrected EEG and EEG uncertainties, which serve as key parameters for measuring the credibility of each modality in the subsequent evidence fusion stage.

[0039] It should be noted that by introducing vehicle behavior stability indicators to calibrate the initial modal uncertainty online, the system can distinguish between physiological changes caused by real fatigue and signal fluctuations caused by driving operations such as rapid acceleration and lane changes. The design of the lightweight calibration network balances computational efficiency and calibration accuracy, ensuring real-time operation on the embedded vehicle platform, while improving the accuracy of uncertainty estimation, laying a reliable foundation for subsequent high-confidence fusion.

[0040] S6. Generate corresponding belief quality based on the modulated EEG evidence vector, modulated EEG evidence vector, calibrated EEG uncertainty, and calibrated EEG uncertainty. Use Dempster-Shafer evidence theory to fuse the belief quality, output the fused fatigue state, and perform graded early warning based on the fused fatigue state.

[0041] Furthermore, regarding the modulated EEG evidence vector Each element is summed and normalized to obtain the EEG support. ,in ; Modulated electrooculography evidence vector Each element is summed and normalized to obtain the electrooculogram support. ,in .

[0042] Calculating EEG uncertainty Electrooculometry uncertainty ; Calculate the evidence conflict coefficient between the two modalities ;like Then, after calculation and fusion, the first... Support for fatigue-like states With fusion uncertainty The formula is as follows: ; ; In the formula, This indicates the degree of support from EEG modalities for the first type of fatigue state. This indicates the degree of support of the electrooculogram (EOG) mode for the type I fatigue state. This represents the prediction uncertainty of EEG modalities. This represents the prediction uncertainty of the electrooculogram (EOG) mode. This indicates the degree of evidence conflict between the two a-modes; the output after fusion is the fatigue state. .

[0043] Set the first uncertainty threshold With the second uncertainty threshold ,satisfy When the fatigue state after fusion is severe fatigue and the fusion uncertainty is... When this occurs, strong intervention measures are triggered, including audible alarms, seat vibration, and red flashing on the dashboard; when the fatigue state after fusion is mild fatigue, or the fatigue state after fusion is severe fatigue but the fusion uncertainty is low... and At that time, only weak prompts are triggered, including voice reminders and yellow prompts on the dashboard; when the fusion uncertainty is... At that time, no warning was triggered.

[0044] It should be noted that by integrating support and uncertainty based on the Dempster-Shafer evidence theory and combining them with the conflict coefficient for adaptive weighting, the confidence of judgment can be enhanced when modal opinions are consistent, and uncertainty can be reasonably retained when modal opinions are conflicting, thus avoiding erroneous decisions caused by blind integration.

[0045] Furthermore, a tiered early warning strategy is implemented based on the fusion uncertainty to ensure that strong interventions are triggered only when there is high confidence and severe fatigue, thus ensuring driving safety, reducing unnecessary disturbances, and improving the intelligence of human-computer interaction and user experience.

[0046] This embodiment also provides a computer device applicable to the driving fatigue recognition method, computer device and storage medium based on EEG-EOG uncertainty evidence fusion, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the driving fatigue recognition method, computer device and storage medium based on EEG-EOG uncertainty evidence fusion as proposed in the above embodiment.

[0047] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0048] This embodiment also provides a storage medium on which a computer program is stored. When executed by a processor, the program implements the driving fatigue recognition method, computer device, and storage medium based on the fusion of uncertain evidence from EEG and EEG as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0049] In summary, this invention establishes a dynamic judgment threshold using awake period data to achieve precise adaptation to the driver's physiological state, introduces contextual information to dynamically suppress the weights of interfering modes, and calibrates modal uncertainty through vehicle behavior indicators to effectively distinguish between true fatigue and pseudo-anomalies. The Dempster-Shafer fusion rule explicitly models the relationship between support, uncertainty, and conflict, and can still output reasonable confidence judgments in high-conflict scenarios to avoid false alarms or missed alarms. Based on fusion uncertainty, it implements graded early warning to ensure that strong intervention is only triggered when there is high confidence and severe fatigue, taking into account both safety performance and user experience. This invention solves the problems of multimodal physiological signals being easily interfered with, having weak generalization ability, and unreliable fusion results in the vehicle environment.

[0050] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for driver fatigue recognition based on the fusion of uncertain evidence from electroencephalography (EEG) and electrooculography (EOG), characterized in that: include: Collect EEG and EEG signals of drivers in a conscious state, construct a personalized physiological baseline containing multidimensional feature statistical parameters, and set a dynamic fatigue judgment threshold based on the personalized physiological baseline. During driving, real-time EEG signals, real-time EOG signals, and contextual information reflecting the current driving environment and vehicle operating status are acquired simultaneously. The prior reliability weights of EEG and EEG modalities are determined based on contextual information. Real-time EEG and EEG signals are preprocessed and depersonalized to obtain EEG feature tensors and EEG feature vectors adapted to individual characteristics. The EEG feature tensor and EEG feature vector are respectively input into the corresponding pre-trained branch model to generate the original EEG evidence vector and the original EEG evidence vector. The original EEG evidence vector and the original EEG evidence vector are modulated by combining the prior reliability weight to obtain the modulated EEG evidence vector and the modulated EEG evidence vector. The initial modal uncertainty is calculated based on the modulated EEG evidence vector and the modulated EEG evidence vector, and the initial modal uncertainty is calibrated online through a lightweight calibration network using the vehicle behavior stability index, outputting the calibrated EEG uncertainty and the calibrated EEG uncertainty. Based on the modulated EEG evidence vector, modulated EEG evidence vector, and calibrated EEG and EEG uncertainties, corresponding belief quality is generated. The Dempster-Shafer evidence theory is used to fuse the belief quality, output the fused fatigue state, and perform graded early warning based on the fused fatigue state.

2. The method for driver fatigue recognition based on EEG-EOG uncertainty evidence fusion as described in claim 1, characterized in that: The construction of a personalized physiological baseline containing multidimensional feature statistical parameters includes the following steps: EEG and EOG signals were continuously collected while the driver was awake and at rest, with a sampling time of no less than five minutes. The EEG signal was segmented and windowed, with each window lasting two seconds and having a 50% overlap. The differential entropy and power spectral density were calculated for each segment. , , , The energy proportion of each frequency band, along with the sample entropy, forms the EEG feature vector; Waveform detection is performed on the electrooculogram (EOG) signal to extract the duration of a single blink, the interval between adjacent blinks, and the slow-phase eye movement velocity, which constitute the EOG feature vector. Calculate the mean of all EEG feature vectors by dimension. with standard deviation Calculate the mean of all electrooculography feature vectors by dimension. with standard deviation These combinations form a personalized physiological baseline; Where represents the mean of the i-th dimension of EEG characteristics in the conscious state, represents the corresponding standard deviation, represents the mean of the i-th dimension of EEG characteristics in the conscious state, and represents the corresponding standard deviation.

3. The method for driver fatigue recognition based on EEG-EOG uncertainty evidence fusion as described in claim 2, characterized in that: The process of setting dynamic fatigue assessment thresholds based on personalized physiological baselines includes the following steps: The first of the EEG characteristics Dimensionally set upper and lower threshold values ​​for the electrooculogram feature. Define upper and lower bound thresholds; When the real-time extracted EEG or EEG features exceed the upper and lower bound thresholds of the corresponding dimension, the dimension is marked as a fatigue-sensitive dimension. The upper and lower bound thresholds are set using a uniform scaling factor. ,satisfy ; EEG characteristics The upper bound threshold of dimension is The lower bound threshold is ; Electroocular characteristics The upper bound threshold of dimension is The lower bound threshold is ; in, This is a preset constant used to control the tolerance bandwidth of an individual baseline.

4. The method for driver fatigue recognition based on EEG-EOG uncertainty evidence fusion as described in claim 3, characterized in that: The determination of prior reliability weights for EEG and EEG modalities based on contextual information includes the following steps: The vehicle obtains current light intensity, vehicle speed, longitudinal acceleration, and lateral acceleration from the vehicle's CAN bus, and calculates the road bump index from the suspension vibration signal; If the light intensity is below the dark light threshold, the prior reliability weight of the electrooculogram modality is set to an attenuation factor of less than one; otherwise, it is set to one. If the road surface bump index is higher than the bump threshold, the weight of the prior reliability of the EEG modality is set to a decay factor of less than one; otherwise, it is set to one. The prior reliability weight is used to modulate the original evidence vector subsequently, reflecting the degree of influence of the external environment on the quality of physiological signals.

5. The driver fatigue recognition method based on EEG-EOG uncertainty evidence fusion as described in claim 4, characterized in that: The depersonalization process includes the following steps: Perform a standardization transformation on each dimension of the real-time EEG feature vector, that is, subtract the corresponding baseline mean from the feature value of that dimension and divide by the corresponding baseline standard deviation to obtain the depersonalized EEG feature tensor; Perform the same standardized transformation on each dimension of the real-time electrooculogram feature vector to obtain a depersonalized electrooculogram feature vector. The standardization transformation maps the feature representations of different drivers to the same relative offset space, eliminating absolute magnitude differences between individuals.

6. The method for driver fatigue recognition based on EEG-EOG uncertainty evidence fusion as described in claim 5, characterized in that: The process of inputting the EEG feature tensor and EEG feature vector into the corresponding pre-trained branch model to generate the original EEG evidence vector and the original EEG evidence vector, and then modulating the original EEG evidence vector and the original EEG evidence vector by combining the prior reliability weights to obtain the modulated EEG evidence vector and the modulated EEG evidence vector, includes the following steps: The depersonalized EEG feature tensor is input into a pre-trained convolutional neural network branch, which outputs the raw EEG evidence vector. ; The depersonalized electrooculogram (EOG) feature vector is input into a pre-trained fully connected neural network branch, which outputs the original EOG evidence vector. ; The original EEG evidence vector and the prior reliability weights of EEG modalities are combined. Multiplying them yields the modulated EEG evidence vector. ; The original electrooculogram evidence vector and the electrooculogram modal prior reliability weights are combined. Multiplying them together yields the modulated electrooculography evidence vector. ; in, This indicates the total number of fatigue state categories. Indicates the EEG branches on the first The strength of the original evidence for fatigue-like states. Indicates the effect of the electrooculography branch on the first The strength of the original evidence for fatigue-like states. and To take values ​​in Real weights for intervals.

7. The method for driver fatigue recognition based on EEG-EOG uncertainty evidence fusion as described in claim 6, characterized in that: The process of generating corresponding belief quality based on modulated EEG evidence vector, modulated EEG evidence vector, calibrated EEG uncertainty, and calibrated EEG uncertainty, fusing the belief quality using Dempster-Shafer evidence theory, and outputting the fused fatigue state includes the following steps: Modulated EEG Evidence Vector Each element is summed and normalized to obtain the EEG support. ,in ; Modulated electrooculo evidence vector Each element is summed and normalized to obtain the electrooculogram support. ,in ; Calculating EEG uncertainty Electrooculometry uncertainty ; Calculate the evidence conflict coefficient between the two modalities ; like Then, after calculation and fusion, the first... Support for fatigue-like states With fusion uncertainty The formula is as follows: ； ； In the formula, Indicates the EEG modality for the first The degree of support for fatigue-like states Indicates the electrooculography mode for the first The degree of support for fatigue-like states This represents the prediction uncertainty of EEG modalities. This represents the prediction uncertainty of the electrooculogram (EOG) mode. This indicates the degree of evidence conflict between the two modalities; the output after fusion is the fatigue state. .

8. The method for driver fatigue recognition based on EEG-EOG uncertainty evidence fusion as described in claim 7, characterized in that: The step of performing graded early warning based on the fatigue state after fusion includes the following steps: Set the first uncertainty threshold With the second uncertainty threshold ,satisfy ; When the fatigue state after fusion is severe fatigue and the fusion uncertainty is... When this occurs, strong intervention measures are triggered, including audible alarms, seat vibration, and red flashing on the dashboard; When the fatigue state after fusion is mild fatigue, or the fatigue state after fusion is severe fatigue but the fusion uncertainty is high. and At this time, only weak prompts are triggered, including voice reminders and yellow prompts on the dashboard; When fusion uncertainty At that time, no warning was triggered.

9. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the driving fatigue recognition method based on EEG-EOG uncertainty evidence fusion as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the driving fatigue recognition method based on EEG-EOG uncertainty evidence fusion as described in any one of claims 1 to 8.

Images