A method for removing driver EEG artifacts in human-machine co-driving

Abstract

The present invention provides a method for removing driver EEG artifacts for human-machine co-driving. The method includes: collecting EEG signals and driving behavior data of the driver simulating driving to obtain original EEG data; identifying various artifact components in the original EEG data, and constructing multiple artifact component combination schemes; performing artifact component removal processing on the EEG data under different artifact component combination schemes, and extracting the spectral features of the EEG signals respectively; calculating the driving behavior data according to the corresponding time period of the spectral features of the EEG signals; evaluating the accuracy of the prediction results of the driving behavior data of the prediction model under different artifact combination removal effects, and obtaining the optimal artifact component combination removal scheme based on the evaluation results. The present invention can clearly define the artifact components that should be removed in different driving scenarios, and while effectively removing the artifacts, it can retain effective information as much as possible, which is helpful for the human-machine co-driving system to use EEG signals to perceive the driving status and make auxiliary decisions.

Description

Technical Field

[0001] The present invention belongs to the field of electroencephalogram (EEG) signal processing and relates to a method for removing EEG artifacts of a driver applied to human-machine co-driving. Background Art

[0002] Electroencephalogram (EEG) signals contain high-frequency neural activity information. EEG signals can not only perceive and monitor the driver's state, ensuring safe driving, but also identify driving intentions by predicting decision-making indicators (acceleration, lane offset, steering wheel angle, etc.), thereby assisting in decision-making and guiding driving behavior. This provides a new approach to perception and decision-making in human-machine co-driving systems. However, EEG signals collected at the scalp are extremely weak. Furthermore, in simulated real-world driving scenarios, EEG signals are affected by the subject's intended and unconscious behaviors, such as body movements, blinks, and saccades, as well as other unexpected interference, such as electronic interference. The strong randomness, nonlinearity, and nonstationarity of EEG signals inevitably lead to various artifacts during the collection process, which in turn affect the accuracy and reliability of correlation analysis. Therefore, removing artifacts from EEG signals is crucial.

[0003] One existing method for removing artifacts from EEG signals involves using ICA (Independent Components Analysis), an independent component extraction method that converts the original EEG signal into independent components using the equation C = WX to identify and remove artifacts. In this equation, X represents the original EEG signal matrix, where rows represent the number of electrode channels and columns represent the time points; C represents the independent component matrix, where rows represent the number of components and columns represent the time points; and W represents the weight matrix, where rows represent the number of components and columns represent the number of electrode channels.

[0004] A drawback of the aforementioned prior art method for removing artifacts from EEG signals is that it relies on operator experience and screens for artifact components by observing EEG spatial scalp maps and the power spectral density of each independent component. This manual visual inspection is time-consuming and labor-intensive, especially when dealing with real-world EEG data, where the fuzzy features of some independent components can make artifact removal even more difficult. While various automatic artifact recognition methods currently exist, setting all identified artifact components to zero does not guarantee a completely clean EEG signal after removal and may also result in the loss of some valid information. Striking a balance between removing contaminated information and losing valuable information is a difficult task.

[0005] Therefore, a method is needed to clearly identify the artifact components that need to be removed in different real-world scenarios in order to obtain the maximum benefit of driving decision-making indicator prediction. Summary of the Invention

[0006] An embodiment of the present invention provides a method for removing driver EEG artifacts applied to human-machine co-driving, so as to effectively assist the human-machine co-driving system in driver status perception and auxiliary decision-making.

[0007] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions.

[0008] A method for removing driver EEG artifacts applied to human-machine co-driving, comprising:

[0009] Collect the driver's EEG signals and driving behavior data while simulating driving, and perform basic noise reduction on the EEG signals to obtain the original EEG data;

[0010] Identify various artifact components in the original EEG data and construct multiple artifact component combination schemes; perform artifact component removal processing on the EEG data under different artifact component combination schemes, and extract the spectral characteristics of the EEG signals after different artifact component combinations are removed;

[0011] Calculating the driver's driving behavior data according to the time period corresponding to the spectrum characteristics of the EEG signal;

[0012] A prediction model is selected based on the spectral characteristics of the EEG signal and the driving behavior data, the accuracy of the prediction results of the driving behavior data of the prediction model under different artifact component removal combinations is evaluated, and the optimal artifact component removal combination scheme is obtained based on the evaluation results.

[0013] Preferably, the collecting of the driver's EEG signals and driving behavior data while simulating driving, and performing basic noise reduction processing on the EEG signals to obtain raw EEG data, includes:

[0014] The driver's EEG signals and driving behavior data of simulated driving are collected synchronously, where the EEG signal sampling frequency is 1000 Hz and the driving behavior data recording frequency is 10 Hz. The collected EEG signals are filtered through a 0.5 Hz to 70 Hz bandpass filter to remove the sinusoidal noise of a specific frequency in the EEG signals, eliminate damaged electrode channels in the EEG signals, and supplement them by interpolation of adjacent electrodes. The EEG data in the EEG signals is set as the average reference to obtain the original EEG data.

[0015] Preferably, the identifying of various artifact components in the original EEG data, constructing multiple artifact component combination schemes, and performing artifact component removal processing on the EEG data under different artifact component combination schemes include:

[0016] After the first round of independent component analysis (ICA) of the raw EEG data, the ADJUST EEG artifact automatic detector based on joint spatial and temporal features was used to identify four artifact components: blinks, horizontal eye movements, vertical eye movements, and general discontinuities. Multiple artifact component combination schemes were formed based on the five artifact components of blinks, horizontal eye movements, vertical eye movements, general and muscle artifacts.

[0017] The EEG data were processed for artifact removal under different artifact component combination schemes. The muscle artifacts were removed by the EMG signal removal function based on blind source separation technology in ARR. The other four artifact components were directly deleted in the ADJUST plug-in.

[0018] Preferably, the extracting of spectral features of the EEG signals after removing different combinations of artifact components comprises:

[0019] The EEG data after removing artifact components were divided into multiple time periods with a length of 1S and an overlapping time of 0.1S. The EEG data of all time periods were combined into a data set, which was divided into a training set and a test set. The EEG data of each time period were converted to different frequency bands by fast Fourier transform, and the average power of the EEG data in five different bands was calculated to obtain the spectral characteristics of the EEG signal. The five different bands included δ band: 0.5–3Hz, θ band: 4–7Hz, α band: 8–13Hz, β band: 14–30Hz and γ band: 31–70Hz.

[0020] Preferably, the calculation of the driver's driving behavior data according to the time period corresponding to the spectrum characteristics of the EEG signal includes:

[0021] Based on the EEG data of each time period, the average value of the driver's driving behavior data within 1 second in each time period is calculated, the EEG data and the corresponding driving behavior data in each time period are stored, and appropriate driving behaviors are selected for specific driving scenarios as driving behavior characteristics and driving decision indicators.

[0022] Preferably, the method of selecting a prediction model based on the spectral characteristics of the EEG signal and the driving behavior data, evaluating the accuracy of the prediction results of the driving behavior data of the prediction model under different artifact component removal combinations, and obtaining the optimal artifact component removal combination scheme based on the evaluation results includes:

[0023] Selecting multiple prediction models from a prediction model library based on the spectral characteristics of the EEG signal and the driving behavior data, wherein each prediction model uses the driving behavior characteristics and the spectral characteristics of the EEG signal as input data, and outputs a predicted driving decision indicator;

[0024] For each prediction model, the mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and R-squared index are calculated on the test set. The calculation formulas are:

[0025]

[0026]

[0027]

[0028] Among them, y i represents the i-th observation value, collected during the driving experiment, represents the i-th predicted value, n is the period, Represents the mean of observations, MAE, RMSE, and MAPER represent the deviation between the predicted value and the observed value. The smaller the value, the lower the prediction error, and the larger the R-square index, the better the model fitting effect;

[0029] Integrate the prediction performance of different prediction models, Oln represents the overall performance index, ln p is the performance indicator of the p-th prediction model, w p is the influence weight of the p-th prediction model;

[0030]

[0031] By comparing the overall performance index of different artifact removal schemes, the artifact removal scheme with the best performance is selected as the artifact removal scheme for this driving scenario.

[0032] It can be seen from the technical solutions provided by the above-mentioned embodiments of the present invention that the present invention can clearly identify the artifact components that should be removed in different driving scenarios, and while effectively removing the artifacts, it can retain effective information as much as possible, which helps the human-machine co-driving system to use EEG signals to perceive the driving status and make auxiliary decisions.

[0033] Additional aspects and advantages of the present invention will be set forth in part in the following description, will become apparent from the following description, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1A processing flow chart of a method for removing driver EEG artifacts in human-machine co-driving provided by an embodiment of the present invention;

[0036] Figure 2 A schematic diagram of the combination of four artifact components provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limiting the present invention.

[0038] It will be understood by those skilled in the art that, unless expressly stated otherwise, the singular forms "a", "an", "said" and "the" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the description of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, "connected" or "coupled" as used herein may include wireless connections or couplings. The term "and / or" used herein includes any unit and all combinations of one or more associated listed items.

[0039] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art in the art to which the present invention pertains. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art and, unless defined as such herein, will not be interpreted in an idealized or overly formal sense.

[0040] To facilitate understanding of the embodiments of the present invention, several specific embodiments will be further explained below with reference to the accompanying drawings. However, each embodiment does not constitute a limitation on the embodiments of the present invention.

[0041] The embodiments of the present invention propose a performance-based artifact removal framework. Specifically, in specific driving scenarios, the EEG signals and driving behavior data of the driver are synchronously collected. By constructing a combined pool of artifact components, for each artifact removal scheme in the pool, the spectral power of different frequency bands is calculated as the intrinsic factor affecting decision-making. Using driving behavior as a stimulus, driving decision indicators are quantitatively predicted. By integrating different prediction models, the artifact removal scheme that achieves the greatest benefit is derived based on performance.

[0042] The present invention constructs an artifact component combination pool, considering five artifact components and two rounds of ICA processing. Finally, a total of 482 artifact combinations are formed for each set of EEG data, with comprehensive coverage, avoiding the accuracy of decision indicator prediction affected by incomplete artifact removal.

[0043] The embodiment of the present invention provides a processing flow of a method for removing driver's EEG artifacts in human-machine co-driving. Figure 1 As shown, the following processing steps are included:

[0044] Step S10: collecting the driver's EEG signals and driving behavior data while simulating driving, and performing basic noise reduction processing on the EEG signals to obtain original EEG data.

[0045] Through a driving simulation experiment in a car-following scenario, the driver's EEG signals and driving behavior data were simultaneously collected. The EEG signal sampling frequency was 1000Hz, and the driving behavior data was recorded at a frequency of 10Hz. The collected EEG signal was bandpass filtered from 0.5Hz to 70Hz to remove sinusoidal noise of specific frequencies in the EEG signal. Damaged electrode channels in the EEG signal were removed and supplemented by interpolation of adjacent electrodes. The EEG (electroencephalogram) data in the EEG signal was set as the average reference to obtain the raw EEG data.

[0046] Step S20: Identify various artifacts in the original EEG data and construct an artifact component combination pool.

[0047] After the first round of ICA (Independent component analysis) on the raw EEG data, ADJUST (An automatic EEG artifact detector based on the joint use of spatial and temporal features) was used to identify four artifact components: blinks, horizontal eye movements, vertical eye movements, and general discontinuities.

[0048] The above four artifact components plus muscle artifacts total five artifact components, which can form a variety of artifact component combinations. In practical applications, it is necessary to selectively remove the above five artifact components. Too much or too little artifact component removal will cause EEG data distortion, thereby affecting the accuracy of subsequent EEG data analysis results. Therefore, the embodiment of the present invention needs to find the best artifact component combination, so that after the EEG data is subjected to artifact component removal processing according to the best artifact component combination, the most accurate driving behavior characteristics of the driver can be analyzed based on the processed EEG data.

[0049] The four types of artifacts identified by ADJUST can form Figure 2 The number of component combinations shown is C14 + C24 + C34 + C44 = 16. This number doubles when muscle artifacts are considered. Similarly, after the second round of ICA, there are 2 × 15 × 15 = 450 new artifact components. Therefore, each EEG data set has a total of 450 + 32 = 482 artifact component combinations, regardless of whether one or two rounds of ICA are performed.

[0050] Step S30: extracting the spectral features of the EEG signal based on the removal results of different artifact combinations.

[0051] The embodiment of the present invention needs to perform artifact component removal processing on the EEG data according to different artifact component combinations. The five different artifacts of blinking, horizontal eye movement, vertical eye movement, general discontinuity and muscle are removed from the above-mentioned original EEG data. Among them, the removal of muscle artifacts is achieved by the electromyography signal removal function based on blind source separation (BSS) technology in ARR. After the other four artifact components are identified through independent component analysis and ADJUST, the corresponding artifact components are directly deleted in the ADJUST plug-in.

[0052] The EEG data after removing artifact components are divided into multiple time periods with a length of 1S and an overlap time of 0.1S. The EEG data of all time periods constitute a data set, which is divided into a training set (80%) and a test set (20%).

[0053] The EEG data of each time period is converted into different frequency bands through fast Fourier transform.

[0054] The average power of EEG data in five different bands (δ band: 0.5–3 Hz, θ band: 4–7 Hz, α band: 8–13 Hz, β band: 14–30 Hz, and γ band: 31–70 Hz) was calculated to obtain the spectral characteristics of the EEG signal.

[0055] Step S40: Calculate the driver's driving behavior data according to the time period corresponding to the spectral characteristics of the EEG signal, and select a prediction model according to the spectral characteristics of the EEG signal and the driving behavior data.

[0056] Based on the EEG data for each time period, calculate the average of driving behavior data within 1 second, including headway, speed difference, velocity, and longitudinal acceleration. Store the EEG data and corresponding driving behavior data (headway, speed difference, velocity, and longitudinal acceleration) for each time period. This example is a car-following driving experiment. In car-following scenarios, headway, speed difference, velocity, and longitudinal acceleration play an important role and can directly reflect driving behavior, so these indicators are selected as driving behavior characteristics.

[0057] Three prediction models, namely, multivariate linear regression, support vector regression, and extra tree regression, were selected from the prediction model library. In the vehicle-following scenario, in each prediction model, the three driving behavior characteristics of headway, speed, and speed difference, and the spectral characteristics of the above-mentioned EEG signals were used as input data. The spectral characteristics of the EEG signals were used as the intrinsic factors affecting decision-making, and the driving behavior characteristics (headway, speed difference, and the mean of speed) were used as stimuli to quantitatively predict longitudinal acceleration, that is, the acceleration along the vehicle's direction of travel, as a driving decision-making indicator in the vehicle-following scenario.

[0058] Longitudinal acceleration, or acceleration along the vehicle's direction of travel, is the primary response factor in vehicle-following scenarios. Therefore, in this example, it is selected as the model's predicted value and driving decision indicator. The predicted result is the longitudinal acceleration value within a single second, corresponding to the observed value, the average longitudinal acceleration per second, calculated above. For other driving scenarios, other driving behavior indicators, such as lane offset and steering wheel angle, can be selected as driving decision indicators and model prediction targets.

[0059] Step S50 : evaluating the accuracy of the prediction results of the driving behavior data of the prediction model under different artifact component combination removal effects, and obtaining the optimal artifact component combination removal solution according to the evaluation results.

[0060] For each prediction model, the mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE) and R-square index are calculated on the test set to evaluate the prediction performance of the prediction model. The formulas are:

[0061]

[0062]

[0063]

[0064]

[0065] Among them, y i represents the i-th observation value, collected during the driving experiment; represents the i-th predicted value; n is the cycle; represents the observed mean.

[0066] For the driving behavior data of longitudinal acceleration, the collected data is used as the observation value, and the prediction result of the model is used as the predicted value. The prediction performance of the prediction model is evaluated by the predicted value and the observed value on the test set.

[0067] For the above four evaluation indicators, the prediction performance of different prediction models is integrated through the following formula. In the formula, Oln represents the overall performance index, ln p is the performance indicator of the p-th prediction model, w p is the influence weight of the p-th prediction model.

[0068]

[0069] The mean values of the above four evaluation indicators (MAE, RMSE, MAPE and R-square index) on the test set were calculated as the performance indicators of each model ln; the difference in the R-square index between the support vector regression and the additional tree regression model and the multiple linear regression model was normalized and used as the influence weight; the overall performance of the support vector regression and the additional tree regression model was integrated by the above formula to obtain its overall performance index on the four evaluation indicators; different artifact component combination schemes were compared, and the four evaluation indicators were comprehensively analyzed to obtain the optimal artifact component combination scheme, among which MAE, RMSE, and MAPE represent the deviation between the predicted value and the observed value. The smaller the value, the lower the prediction error, and the larger the R-square index, the better the model fitting effect.

[0070] Here, for each evaluation indicator (MAE, RMSE, MAPE and R-squared index), the prediction performance of different prediction models is integrated through the above-mentioned Oln calculation formula, that is, for each artifact removal scheme, its overall performance index in MAE, RMSE, MAPE and R-squared index is obtained respectively.

[0071] By comparing the overall performance index of different artifact removal schemes, the artifact removal scheme with the best performance is selected as the artifact removal scheme for this driving scenario.

[0072] According to the above steps, in the vehicle-following driving scenario, the artifact removal solution obtained by the method of the present invention is better than removing all identified artifacts, showing excellent denoising capabilities compared to traditional methods in specific driving scenarios.

[0073] In summary, the driver EEG artifact removal framework in the embodiment of the present invention is based on performance. The framework can clearly identify the specific artifact components that should be removed in different driving scenarios. The existing technology usually sets all identified artifacts to zero, ignoring the influence of valid information contained in the removed components; the artifact component combination pool constructed by the present invention contains five artifact components and two rounds of ICA processing, while the existing technology often ignores the problem that artifact components still exist after one ICA.

[0074] The method provided by the present invention solves the problems in the prior art of unclear artifact components to be removed and loss of effective EEG information during artifact removal. The processed EEG signals can be better applied to driving status perception and auxiliary decision-making of human-machine co-driving systems.

[0075] The driver EEG artifact removal method for human-machine co-driving, provided by an embodiment of the present invention, constructs a combined pool of artifact components encompassing five artifact components and two rounds of ICA. This method, from a performance perspective, identifies the most effective artifact removal solution for different driving scenarios, effectively removing artifact components while preserving as much valid information as possible. The processed EEG signals can be better applied to predict decision-making indicators, assisting the human-machine co-driving system in driver status perception and decision-making support.

[0076] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of an embodiment, and the modules or processes in the accompanying drawings are not necessarily required to implement the present invention.

[0077] From the above description of the embodiments, it can be seen that those skilled in the art can clearly understand that the present invention can be implemented by means of software plus the necessary general-purpose hardware platform. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present invention or certain parts of the embodiments.

[0078] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, refer to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without making any creative efforts.

[0079] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for removing driver's EEG artifacts in human-machine co-driving, characterized in that: include: Collect the driver's EEG signals and driving behavior data while simulating driving, and perform basic noise reduction on the EEG signals to obtain the original EEG data; Identify various artifact components in the original EEG data and construct multiple artifact component combination schemes; perform artifact component removal processing on the EEG data under different artifact component combination schemes, and extract the spectral characteristics of the EEG signals after different artifact component combinations are removed; Calculating the driver's driving behavior data according to the time period corresponding to the spectrum characteristics of the EEG signal; selecting a prediction model based on the spectral characteristics of the EEG signal and the driving behavior data, evaluating the accuracy of the prediction results of the driving behavior data of the prediction model under different artifact component removal combinations, and obtaining an optimal artifact component removal combination scheme based on the evaluation results; The aforementioned identification of various artifact components in the original EEG data, construction of multiple artifact component combination schemes, and performing artifact component removal processing on the EEG data under different artifact component combination schemes include: After the first round of independent component analysis (ICA) of the raw EEG data, the ADJUST EEG artifact automatic detector based on joint spatial and temporal features was used to identify four artifact components: blinks, horizontal eye movements, vertical eye movements, and general discontinuities. Multiple artifact component combination schemes were formed based on the five artifact components of blinks, horizontal eye movements, vertical eye movements, general and muscle artifacts. Find the best combination of artifact components and perform artifact removal on the EEG data according to this best combination. Muscle artifacts are removed using the EMG signal removal function based on blind source separation technology in ARR. The other four artifact components are directly deleted in the ADJUST plug-in. The said extracting the spectrum features of the EEG signal after removing different combinations of artifact components includes: The EEG data after removing artifact components were divided into multiple time periods with a length of 1s and an overlap time of 0.1s. The EEG data of all time periods were combined into a data set, which was divided into a training set and a test set. The EEG data of each time period were converted to different frequency bands by fast Fourier transform. The average power of the EEG data of five different bands was calculated to obtain the spectral characteristics of the EEG signal. The five different bands included delta band: 0.5–3Hz, theta band: 4–7Hz, alpha band: 8–13Hz, beta band: 14–30Hz, and gamma band: 31–70Hz. The method of calculating the driver's driving behavior data according to the time period corresponding to the spectrum characteristics of the EEG signal includes: Based on the EEG data of each time period, the average value of the driver's driving behavior data within 1 second in each time period is calculated, the EEG data and the corresponding driving behavior data in each time period are stored, and appropriate driving behaviors are selected for specific driving scenarios as driving behavior characteristics and driving decision indicators.

2. The method according to claim 1, characterized in that The process of collecting the driver's simulated driving EEG signals and driving behavior data and performing basic noise reduction processing on the EEG signals to obtain raw EEG data includes: The driver's EEG signals and driving behavior data of simulated driving are collected synchronously, where the EEG signal sampling frequency is 1000 Hz and the driving behavior data recording frequency is 10 Hz. The collected EEG signals are filtered through a 0.5 Hz to 70 Hz bandpass filter to remove the sinusoidal noise of a specific frequency in the EEG signals, eliminate damaged electrode channels in the EEG signals, and supplement them by interpolation of adjacent electrodes. The EEG data in the EEG signals is set as the average reference to obtain the original EEG data.

3. The method according to claim 1 or 2, characterized in that The method of selecting a prediction model based on the spectral characteristics of the EEG signal and the driving behavior data, evaluating the accuracy of the prediction results of the driving behavior data of the prediction model under different artifact component removal combinations, and obtaining the optimal artifact component removal solution based on the evaluation results includes: Selecting multiple prediction models from a prediction model library based on the spectral characteristics of the EEG signal and the driving behavior data, wherein each prediction model uses the driving behavior characteristics and the spectral characteristics of the EEG signal as input data, and outputs a predicted driving decision indicator; For each prediction model, the mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and R-squared index are calculated on the test set. The calculation formulas are: Among them, y i represents the i-th observation value, collected during the driving experiment, represents the i-th predicted value, n is the period, Represents the mean of observations, MAE, RMSE, and MAPER represent the deviation between the predicted value and the observed value. The smaller the value, the lower the prediction error, and the larger the R-square index, the better the model fitting effect; Integrate the prediction performance of different prediction models, Oln represents the overall performance index, ln p is the performance indicator of the p-th prediction model, w p is the influence weight of the p-th prediction model; By comparing the overall performance index of different artifact removal schemes, the artifact removal scheme with the best performance is selected as the artifact removal scheme for this driving scenario.

Images