An improved method for the PVT experimental paradigm based on behavioral and EEG characteristics

Abstract

A method for improving a PVT experimental paradigm based on behavioral and EEG characteristics, the method comprising: changing the visual stimulation form of the PVT experiment, testing subjects, analyzing experimental results, selecting a PVT experimental scheme with the best visual stimulation form for inducing mental fatigue, changing the test duration of the PVT experiment, testing, analyzing experimental results, selecting a PVT experimental scheme with the best test duration for inducing mental fatigue, changing the random stimulation interval of the PVT experiment, testing, analyzing experimental results, and selecting a PVT experimental scheme with the best mental fatigue inducing effect. The present invention solves the problem that the influence of a single factor on the mental fatigue inducing effect of the PVT task paradigm is still unclear. By comparing the differences between different PVT design elements in terms of task performance and EEG characteristics, a PVT design scheme with higher mental fatigue inducing efficiency is found, providing a better experimental method for applications such as mental fatigue detection using machine learning.

Description

Technical Field

[0001] The present invention relates to the field of cognitive neuroscience technology, and in particular to a method for improving a PVT experimental paradigm based on behavior and EEG characteristics. Background Art

[0002] Mental fatigue (also known as brain fatigue) can lead to a decrease in people's alertness and concentration levels, often manifesting as an inability or inability to continue working. Specific conditions include incorrect decisions about work content, inattention, or a decline in working memory performance.

[0003] According to experimental psychologists, mental fatigue can be measured using a simple SR model (stimulus-response model). This approach assumes that the presentation of a stimulus will elicit similar responses in all individuals, or that the same type of response will be observed within a certain timeframe. As early as the 1940s, the British Air Force designed a mental fatigue detection task (the Mackworth Clock Test) to recruit suitable personnel for combat radar surveillance. With the development of experimental methods based on the SR model, the PVT (Psychomotor Vigilance Test), proposed in 1985, has become widely used due to its high sensitivity to mental fatigue states (particularly low alertness) and its indifference to individual ability and learning effects. Compared to experimental paradigms such as the Mackworth Clock Test, which require at least two hours, the PVT's typical test duration of only 10 minutes is a significant factor contributing to its popularity. The PVT measures a subject's ability to sustain attention (i.e., vigilance) by recording their reaction time (RT) to visual (or auditory) stimuli presented at random interstimulus intervals (ISIs). Different from the simplest SR model, PVT not only depends on the specific stimulus type (usually visual) and response method (usually key pressing), but is also affected by the duration of the task and the design of ISI parameters.

[0004] Electroencephalogram (EEG) signals, derived from inhibitory and excitatory postsynaptic potentials (PSPs) in cortical neurons, are widely considered to be the hallmark of mental fatigue. They can couple intrinsic neuronal bioelectrical activity with observable task performance in the spatiotemporal and frequency domains, whereas results from simple psychological experiments are limited to behavioral task performance. Regarding the frequency domain characteristics of mental fatigue-induced EEG, most studies have shown that the power of the α, θ, and β bands varies significantly with fatigue status. In addition to single-band power, researchers have also found that power ratios such as (θ+α / β) and (θ+α / δ) are strongly correlated with mental fatigue. Regarding the temporal characteristics of mental fatigue-induced EEG, numerous studies have demonstrated that, when conducted during the PVT task, the P1 amplitude of the ERP in some channels in the occipital lobe decreases with increasing mental fatigue.

[0005] The PVT, based on visual stimulation and keystroke responses, is a promising and easily implemented model for inducing SR with mental fatigue. However, the PVT task paradigm involves numerous design factors that can be varied, and the impact of each individual factor on the effectiveness of inducing mental fatigue in PVT experiments remains unclear. In addition to primary factors such as stimulus type, response mode, test duration, and inter-stimulus interval, numerous secondary factors (such as hardware model, software platform, and programming method) can influence the performance of PVT-induced mental fatigue. These factors vary significantly between different versions of the PVT, resulting in a low degree of standardization and inconsistent results. Summary of the Invention

[0006] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to propose a coal mine water hazard early warning method based on Spark. By comparing the differences between different PVT design elements in task performance and EEG characteristics, a PVT design scheme with higher mental fatigue induction efficiency is found, providing a better experimental method for applications such as mental fatigue detection using machine learning.

[0007] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A method for improving the PVT experimental paradigm based on behavioral and EEG characteristics, the method comprising:

[0009] Changing the visual stimulation form of the PVT experiment to obtain several groups of PVT experimental schemes with different visual stimulation forms, testing subjects using the several groups of PVT experimental schemes with different visual stimulation forms, obtaining several groups of experimental results, analyzing the several groups of experimental results respectively, and selecting the PVT experimental scheme with the best visual stimulation form that induces mental fatigue;

[0010] Based on the optimal visual stimulation form PVT experimental scheme, the test duration of the PVT experiment is changed to obtain several groups of PVT experimental schemes with different test durations, and the subjects are tested using the several groups of PVT experimental schemes with different test durations to obtain several groups of experimental results. The several groups of experimental results are analyzed separately, and the PVT experimental scheme with the best test duration for inducing mental fatigue is selected;

[0011] Based on the PVT experimental plan with the optimal test duration, the random stimulation interval of the PVT experiment was changed to obtain several groups of PVT experimental plans with different random stimulation intervals. The subjects were tested using several groups of PVT experimental plans with different random stimulation intervals to obtain several groups of experimental results. These groups of experimental results were analyzed separately, and the PVT experimental plan with the best mental fatigue inducing effect was selected.

[0012] The analysis of several groups of experimental results respectively includes:

[0013] The fatigue trial ratio of several groups of experimental results was analyzed respectively, and the PVT experimental scheme with the best fatigue trial ratio was selected;

[0014] Conduct fatigue time-frequency domain analysis on several groups of experimental results and select the optimal PVT experimental scheme in fatigue time-frequency domain;

[0015] Based on the PVT experimental scheme with the best proportion of fatigue trials and the best fatigue time-frequency domain, the PVT experimental scheme with the best visual stimulation form that induces mental fatigue / the PVT experimental scheme with the best test duration that induces mental fatigue / the PVT experimental scheme with the best mental fatigue induction effect are selected.

[0016] The fatigue trial ratio analysis is performed on several groups of experimental results, and the PVT experimental scheme with the best fatigue trial ratio is selected, including:

[0017] Collect behavioral task performance data from each trial during each set of experiments;

[0018] Clean and filter behavioral task performance data to obtain behavioral data segments and local reaction times;

[0019] The behavioral data segments and local reaction time were used to calculate the proportion of fatigue trials in each group of experiments, and the PVT experimental plan with the optimal proportion of fatigue trials was selected.

[0020] The fatigue time-frequency domain analysis is performed on several groups of experimental results, and the optimal PVT experimental scheme in the fatigue time-frequency domain is selected, including:

[0021] Collect the original physiological EEG signals generated in each trial during each experimental session;

[0022] Preprocessing the physiological original EEG signal to obtain the EEG signal;

[0023] Conduct frequency domain feature analysis on the EEG signals of each experimental group and select the PVT experimental scheme with the best frequency domain analysis;

[0024] Perform time domain feature analysis on the EEG signals of each experimental group and select the PVT experimental scheme with the best time domain analysis;

[0025] The optimal PVT experimental scheme in frequency domain analysis and the optimal PVT experimental scheme in time domain analysis are comprehensively analyzed, and the optimal PVT experimental scheme in fatigue time-frequency domain is selected.

[0026] The calculation of the fatigue trial ratio of each group of experiments using behavioral data segments and local reaction time includes:

[0027] For each set of experiments, one trial among several trials, as well as 5 trials before and after, were selected, for a total of 11 trials;

[0028] The local reaction time of 11 trials is obtained, and the local reaction time is expressed as:

[0029] RT=T R -T S ,

[0030] Where RT represents the local reaction time, T S represents the moment when the visual stimulus is presented on the computer screen, T R The moment when the signal indicating that the subject pressed the designated key was received by the computer;

[0031] The global reaction time of each of the 11 trials was calculated using the equal weighted average method. The global reaction time is expressed as:

[0032] GlobalRT i =w i-l RT i-l +…+w i RT i +…+w i+l RT i+l ,

[0033] Where w represents the weight, RT represents the local reaction time, i represents the i-th trial, and l represents the length of the window, that is, the number of adjacent trials included in the calculation;

[0034] The 5th percentile of the local reaction time of 11 trials was used as the benchmark, and 1.25 times the benchmark was taken as the threshold;

[0035] The local reaction time and global reaction time of the 11 trials are compared with the threshold value. If the local reaction time and global reaction time of the current trial are both less than the threshold value, the current trial is judged as an alert trial; if the local reaction time and global reaction time of the current trial are both greater than the threshold value, the current trial is judged as a fatigue trial.

[0036] The proportion of fatigue trials was obtained based on the ratio of fatigue trials to the 11 trials.

[0037] The frequency domain feature analysis of the EEG signals of each group of experiments is performed, and the optimal PVT experimental scheme is selected according to the frequency domain analysis, including:

[0038] The EEG signals of each group of experiments were divided into five frequency bands: δ (1-4 Hz), θ (4-8 Hz), α (8-14 Hz), β (14-31 Hz), and γ (31-50 Hz).

[0039] For each set of experiments, the average logarithmic power of fatigue trials and alert trials in the five frequency bands was calculated respectively to obtain the average logarithmic power of fatigue trials and the average logarithmic power of alert trials. The average logarithmic power is expressed as:

[0040]

[0041] Where N is the number of sampling points, and x[n] is the expression of the discrete signal;

[0042] The differences between the average logarithmic power of fatigue trials and the average logarithmic power of alert trials in the five frequency bands were calculated respectively, and the average logarithmic power difference of the δ frequency band, the θ frequency band, the α frequency band, the β frequency band, and the γ frequency band were obtained.

[0043] The average logarithmic power differences of the five frequency bands in each group of experiments were compared, and the experiments with high average logarithmic power differences in the θ band, high average logarithmic power differences in the α band, and low average logarithmic power differences in the β band were selected as the optimal PVT experimental scheme for frequency domain analysis.

[0044] The time domain feature analysis of the EEG signals of each group of experiments is performed, and the optimal PVT experimental scheme for the time domain analysis is selected, including:

[0045] Obtain the OZ channel ERP waveforms of fatigue trials and alertness trials in the occipital lobe of each experimental group;

[0046] The ERP waveforms of the OZ channel in the fatigue trial and the ERP waveforms of the OZ channel in the alert trial of each group of experiments were compared, and the experiment in which the P1 amplitude of the ERP waveform of the OZ channel in the fatigue trial was smaller than the P1 amplitude of the ERP waveform of the OZ channel in the alert trial was selected as the optimal PVT experimental plan for time domain analysis.

[0047] The preprocessing of the physiological raw EEG signal includes: deleting useless segments, deleting useless electrodes, electrode positioning, low-pass filtering, high-pass filtering, removing power frequency interference, electrode re-reference, reducing the sampling rate, removing obvious bad segments and bad tracks, removing artifacts using ICA technology, band-pass filtering of frequency bands, extracting ERP segments, and performing baseline correction on the ERP segments.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. Based on existing research results, this paper combines local reaction time and global reaction time to propose the concepts of potential reaction time and the corresponding alert reaction time (fatigue threshold). Based on these basic task performance indicators, this paper calculates another metric, the proportion of fatigue trials. Finally, this paper applies these indicators to the behavioral analysis and evaluation of actual experimental fragments, helping to determine whether individual trials in the SR model are fatigued. This can be extended to other research.

[0050] 2. The present invention uses the aforementioned task performance indicators to screen reliable fatigue and alertness trials, and extracts and analyzes the features of the electroencephalographic (EEG) signals of these two types of samples. It finds that the neural activity phenomena in most of the screened experimental segments can be supported by relevant conclusions in the field, that is, they are consistent with the "fatigue time-frequency domain model" proposed by the present invention. Both behavioral and physiological evidence verify the rationality of the present invention's judgment of fatigue trials, thereby providing an evaluation basis for the mental fatigue-inducing effect of the experimental method.

[0051] 3. The present invention discovered that the currently used PVT paradigm can be improved in terms of visual stimulation factors by changing the stimulus form from dynamic numbers to static simple graphics, or eliminating the timely feedback process for each trial. It also verified the widely believed advantages of long-term testing. It was also found that appropriately shortening the random stimulation interval does not lead to a significant decrease in experimental results. On the contrary, it can help experimenters compress experimental time or collect more trials, which is a reasonable alternative to the standard ISI.

[0052] In summary, by comparing the differences in task performance and EEG characteristics of different PVT design elements, the present invention finds a PVT design scheme with higher mental fatigue induction efficiency, providing a better experimental method for applications such as mental fatigue detection using machine learning. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The figure is a flow chart of the design of the mental fatigue inducing experimental method of the present invention.

[0054] Figure 2 Schematic diagram of four reaction times in the PVT task of the present invention.

[0055] Figure 3 This is the brain topography map of the average logarithmic power difference of five frequency bands under fatigue and alertness states of the present invention.

[0056] Figure 4 ERP diagrams of the OZ channel under fatigue and alertness states of the present invention.

[0057] Figure 5 Schematic diagram of the benchmark PVT design scheme of the present invention. DETAILED DESCRIPTION

[0058] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0059] The present invention has the following purposes:

[0060] 1. From the perspective of external behavioral information acquisition, design task performance indicators to evaluate the mental fatigue-inducing effects of experimental methods.

[0061] 2. From the perspective of intrinsic physiological index detection, extract the EEG signal characteristics to evaluate the mental fatigue inducing effect of the experimental method.

[0062] 3. Based on behavioral and physiological evidence, find the visual stimulation form, test duration, random stimulation interval and other factors that are more effective in inducing mental fatigue in the SR model.

[0063] 4. Taking all factors into consideration, we will ultimately determine an experimental method that has a good effect in inducing mental fatigue in appropriate situations.

[0064] 5. Provide experimental paradigm standards that are more accurate, efficient, and easier to conduct quantitative evaluations for data collection and other tasks related to mental fatigue detection using machine learning.

[0065] The present invention is based on the popular SR model experimental paradigm of PVT. At the same time, it relies on behavioral task performance indicators and physiological EEG signal characteristics to select design factors that can effectively improve and enhance the mental fatigue induction effect from the benchmark PVT experimental paradigm, and explores a method suitable for evaluating the mental fatigue induction effect of SR experimental paradigms. Specifically, the present invention proposes a "fatigue trial ratio" indicator from a behavioral perspective to evaluate the mental fatigue induction effect of SR experimental methods; and at the same time, from a physiological perspective, it comprehensively considers the frequency domain and time domain characteristics of EEG signals, and uses a proposed "fatigue time-frequency domain pattern" indicator to combine with behavioral indicators to double verify the effect of the experimental method; through a series of comparative experiments based on PVT design factors, the implementation details worthy of improvement in the PVT paradigm are determined (for example, using static simple graphics as visual stimuli, canceling the timely feedback process, and appropriately reducing the random stimulation interval, etc.), and obtains an experimental paradigm with better mental fatigue induction effect based on PVT.

[0066] The idea of ​​the present invention is:

[0067] 1. Recruit subjects to test the baseline experimental method and record behavioral task performance and physiological EEG signals generated simultaneously during the experiment.

[0068] 2. Clean and filter the task performance data to extract the fragments and reaction time indicators required for subsequent analysis.

[0069] 3. Perform multiple preprocessing operations on the original EEG signal to obtain a relatively clean EEG signal, preparing for subsequent feature extraction.

[0070] 4. Analyze behavioral data and design task performance indicators suitable for evaluating the mental fatigue-inducing effects of experimental methods. Introduce the factor of "proportion of fatigue trials" and use it together with indicators such as average reaction time to analyze and evaluate the collected experimental fragments.

[0071] 5. Extract frequency domain features (logarithmic power) and time domain features (ERP waveform) from the processed EEG signals. Based on existing research on power changes in each frequency band and channel of the EEG and changes in the P1 amplitude of the ERP in the occipital lobe region under fatigue conditions, determine EEG signal features suitable for evaluating the fatigue-inducing effect of the experimental method, namely the "fatigue time-frequency domain pattern." Compare the conclusions with the EEG features extracted from the experimental footage to determine their applicability. The fatigue-inducing effect of the experimental method will then be analyzed and evaluated based on established neural markers of mental alertness.

[0072] 6. Taking the PVT experimental paradigm commonly used in many mental fatigue-induced SR models as a benchmark, based on the behavioral and physiological data characteristics extracted under the above method, by controlling variables for comparison, the influence of various inducing factors in the stimulation (including visual stimulation form, test duration and random stimulation interval, etc.) on the mental fatigue-inducing effect of the experimental method was explored.

[0073] 7. Combine the dual fatigue judgment criteria of behavioral task performance and physiological EEG characteristics to ultimately determine an experimental method that has a good effect in inducing mental fatigue in appropriate situations.

[0074] An improved method of the PVT experimental paradigm based on behavioral and EEG characteristics, such as Figure 1 As shown, the method includes:

[0075] S1, changing the visual stimulation form of the PVT experiment to obtain several groups of PVT experimental schemes with different visual stimulation forms, testing subjects using several groups of PVT experimental schemes with different visual stimulation forms, obtaining several groups of experimental results, analyzing the several groups of experimental results respectively, and selecting the PVT experimental scheme with the best visual stimulation form that induces mental fatigue;

[0076] S2, based on the optimal visual stimulation form PVT experimental scheme, changing the test duration of the PVT experiment to obtain several groups of PVT experimental schemes with different test durations, testing subjects using several groups of PVT experimental schemes with different test durations, obtaining several groups of experimental results, analyzing the several groups of experimental results separately, and selecting the PVT experimental scheme with the best test duration for inducing mental fatigue;

[0077] S3. Based on the optimal test duration PVT experimental scheme, change the random stimulation interval of the PVT experiment to obtain several groups of PVT experimental schemes with different random stimulation intervals. Use several groups of PVT experimental schemes with different random stimulation intervals to test the subjects and obtain several groups of experimental results. Analyze the several groups of experimental results separately and select the PVT experimental scheme with the best mental fatigue inducing effect.

[0078] The analysis of several groups of experimental results respectively includes:

[0079] S4, analyze the fatigue trial ratio of several groups of experimental results respectively, and select the PVT experimental plan with the best fatigue trial ratio, specifically:

[0080] S4-1, collect behavioral task performance data generated in each trial during each set of experiments;

[0081] S4-2, clean and filter behavioral task performance data to obtain behavioral data segments and indicators such as local reaction time;

[0082] S4-3, using behavioral data segments and local reaction time, calculate the proportion of fatigue trials in each experimental group, and select the PVT experimental plan with the optimal proportion of fatigue trials. That is, the higher the proportion of fatigue trials, the optimal proportion of fatigue trials. The calculation method of the fatigue trial proportion is:

[0083] For each set of experiments, one trial among several trials, as well as 5 trials before and after, were selected, for a total of 11 trials;

[0084] The local reaction time of 11 trials is obtained, and the local reaction time is expressed as:

[0085] RT=T R -T S ,

[0086] Where RT represents the local reaction time, T S represents the moment when the visual stimulus is presented on the computer screen, T R The moment when the signal indicating that the subject pressed the designated key was received by the computer;

[0087] The global reaction time of each of the 11 trials was calculated using the equal-weighted average method to better reflect the changing trend of the subject's reaction time during the test. For a trial, its local reaction time is the actual reaction time directly measured, while the global reaction time is the weighted sum of the reaction time of the trial and multiple adjacent trials. The global reaction time is expressed as:

[0088] GlobalRT i =wi-l RT i-l +…+w i RT i +…+w i+l RT i+l ,

[0089] Where w represents the weight, RT represents the local reaction time, i represents the i-th trial, and l represents the length of the window, that is, the number of adjacent trials included in the calculation;

[0090] The 5th percentile of the local reaction time of the 11 trials was used as the benchmark, and 1.25 times the benchmark was taken as the threshold for distinguishing fatigue from alertness trials;

[0091] The local reaction time and global reaction time of the 11 trials are compared with the threshold value. If the local reaction time and global reaction time of the current trial are both less than the threshold value, the current trial is judged as an alert trial; if the local reaction time and global reaction time of the current trial are both greater than the threshold value, the current trial is judged as a fatigue trial.

[0092] The proportion of fatigue trials was obtained based on the ratio of fatigue trials to the 11 trials.

[0093] like Figure 2 As shown, the green line represents the local reaction time measured directly, the blue line represents the global reaction time obtained by sliding average, the yellow straight line is the potential reaction time that reflects the reaction time limit of the subject during this test to a certain extent, and the purple straight line is the alert (or fatigue) reaction time as a fatigue distinction standard. Figure 2 It can be found that after completing approximately 30% of trials, the subject's local and global reaction times begin to essentially remain above the fatigue reaction time threshold. Approximately half of the trials meet the aforementioned fatigue trial determination, demonstrating that the experiment successfully induced mental fatigue in the subject at this point. Therefore, the present invention proposes the concept of "the proportion of fatigue trials to total trials" and uses it as a basis for evaluating the induction effect in mental fatigue induction experiments, demonstrating promising results.

[0094] In summary, the present invention uses the proportion of fatigue trials during the experiment as the most important task performance evaluation criterion, and integrates traditional indicators such as average reaction time to jointly achieve the evaluation of the mental fatigue-inducing effect of the experimental method at the behavioral level.

[0095] S5, perform fatigue time-frequency domain analysis on several groups of experimental results, and select the optimal PVT experimental scheme in fatigue time-frequency domain, specifically:

[0096] S5-1, collect the physiological original EEG signals generated in each trial during each experimental group;

[0097] S5-2, preprocessing the original physiological EEG signals to obtain electroencephalographic signals, wherein the preprocessing of the original physiological EEG signals includes: deleting useless segments, deleting useless electrodes, electrode positioning, low-pass filtering, high-pass filtering, removing power frequency interference, electrode re-reference, reducing the sampling rate, removing obvious bad segments and bad tracks, removing artifacts using ICA technology, band-pass filtering, extracting ERP (event-related potential) segments, and performing baseline correction on the ERP segments.

[0098] S5-3, perform frequency domain feature analysis on the EEG signals of each experimental group and select the PVT experimental scheme with the best frequency domain analysis, specifically:

[0099] S5-31, the EEG signals of each experimental group were divided into five frequency bands, including: δ (1-4 Hz), θ (4-8 Hz), α (8-14 Hz), β (14-31 Hz), and γ (31-50 Hz);

[0100] S5-32, for each set of experiments, calculate the average logarithmic power of fatigue trials and alert trials in the five frequency bands, and obtain the average logarithmic power of fatigue trials and the average logarithmic power of alert trials. The average logarithmic power is expressed as:

[0101]

[0102] Where N is the number of sampling points, x[n] is the expression of the discrete signal, and the natural logarithm of the power is used to convert its multiplication relationship into an addition relationship, weaken the influence of extreme values, and reduce the individual differences between different data. In the 59 EEG channels used in the present invention, the difference between the average value of the logarithmic power of each frequency band of the fatigue sample and the alert sample is taken, and a set of five frequency band brain topography maps can be drawn, such as Figure 3 shown.

[0103] S5-33, calculate the difference between the average logarithmic power of fatigue trials and the average logarithmic power of alert trials in the five frequency bands, and obtain the average logarithmic power difference of the δ frequency band, the θ frequency band, the α frequency band, the β frequency band, and the γ frequency band;

[0104] S5-34, compare the average logarithmic power difference of the five frequency bands in each group of experiments, and select the experiment with the highest average logarithmic power difference in the θ band, the highest average logarithmic power difference in the α band, and the lowest average logarithmic power difference in the β band as the optimal PVT experimental scheme for frequency domain analysis. Figure 3 In the figure, warm colors represent positive values ​​(i.e., the average log power of fatigued trials is higher), and cool colors represent negative values ​​(i.e., the average log power of fatigued trials is lower). Figure 3The fatigue trials in this data segment showed overall higher theta band power (in the frontal, central, and occipital regions) and alpha band power (in the central, parietal, and occipital regions), as well as lower beta band power (in the temporal region), consistent with the conclusions of relevant literature. This demonstrates that the experimental method reflected in this data segment has a good effect in inducing mental fatigue, enabling good temporal coupling of neuronal bioelectrical activity and behavioral performance of the task.

[0105] S5-4: Time-domain feature analysis of the EEG signals from each experimental group was performed, and the PVT experimental protocol with the optimal time-domain analysis was selected. In the time domain, the present invention primarily analyzed the event-related potentials (ERPs) in the occipital lobe region of the two sample segments. ERPs are small electrical potentials formed when the brain responds to certain events (or stimuli), reflecting the overall activity of neurons in information processing.

[0106] Specifically:

[0107] S5-41, obtain the OZ channel ERP waveforms of fatigue trials and alertness trials in the occipital lobe area of ​​each group of experiments;

[0108] S5-42, compare the ERP waveforms of the OZ channel in the fatigue trial and the ERP waveforms of the OZ channel in the alert trial in each group of experiments, and select the experiment in which the P1 amplitude of the ERP waveform of the OZ channel in the fatigue trial is smaller than the P1 amplitude of the ERP waveform of the OZ channel in the alert trial as the optimal PVT experimental plan for time domain analysis.

[0109] Figure 4 The red line in the middle represents the OZ channel ERP waveform of the fatigue trial, while the blue line represents the corresponding waveform of the alertness trial. Figure 4 The results show that subjects had an ERP waveform with a lower P1 amplitude during fatigue trials, consistent with findings in relevant literature. These results demonstrate, both behaviorally and physiologically, that the experimental method used in this data effectively induced mental fatigue.

[0110] In summary, by comprehensively considering the power changes of each channel in each frequency band of EEG and the P1 amplitude changes of the ERP waveform of the occipital lobe channel under fatigue state, we can determine an EEG signal feature suitable for evaluating the mental fatigue inducing effect of the experimental method, namely the "fatigue time-frequency domain pattern".

[0111] S5-5, comprehensively consider the optimal PVT experimental scheme in frequency domain analysis and the optimal PVT experimental scheme in time domain analysis, and select the optimal PVT experimental scheme in fatigue time-frequency domain.

[0112] S6, based on the PVT experimental scheme with the best proportion of fatigue trials and the best fatigue time-frequency domain, select the PVT experimental scheme with the best visual stimulation form that induces mental fatigue / the PVT experimental scheme with the best test duration that induces mental fatigue / the PVT experimental scheme with the best mental fatigue induction effect.

[0113] The PVT experimental scheme with the best mental fatigue inducing effect selected by the present invention was verified:

[0114] This paper mainly studies the improvement methods of design factors such as visual stimulation form, test duration and random stimulation interval in the PVT paradigm, aiming to determine the relatively optimal experimental implementation method.

[0115] From the perspective of visual stimulation form, the factors that can be studied include the color of the background of stimulus presentation, which is generally neutral gray and pure black; the shape and color of the fixation point presented at the stimulus interval, which can generally be a white cross or square; the form and color of the stimulus, which are generally selected from static simple graphics and dynamic scrolling numbers, and the choice of color generally follows the idea of ​​high contrast relative to the background; and the choice and method of feedback. It is not necessary to promptly show the subject the feedback step of the current trial reaction time, and its presentation form and color can also change.

[0116] When considering the duration of the test, it is generally believed that shorter PVT tasks are difficult to fully induce the subject's mental fatigue, but longer PVT tasks will significantly increase the difficulty of implementing the experimental method. For example, the subject may experience negative emotions such as boredom due to long-term testing.

[0117] Furthermore, the randomized interstimulus interval setting is one of the key differences between PVT and simple SR models. A common setting is to evenly distribute the interstimulus intervals between 2 and 10 seconds. Studies have shown that reducing the randomized interstimulus interval improves subjects' ability to focus, leading to better overall task performance, with faster average reaction times. While shorter randomized interstimulus intervals may not induce mental fatigue in subjects, this approach allows subjects to complete more trials within the same timeframe, facilitating the construction of machine learning datasets.

[0118] First, the most widely used PVT design scheme is used as a benchmark, and then the influence of each design factor on the mental fatigue induction effect of the experimental method is explored one by one by controlling the variables. In terms of visual stimulation form, the benchmark scheme uses a black stimulus presentation background, a red square-shaped fixation point, a yellow dynamic digital stimulus in a red square, and a yellow digital feedback in a yellow square; in terms of test duration, the scheme is designed to be suitable for a typical value of 10 minutes; in terms of random stimulus intervals, the scheme is set to be evenly distributed within 2 to 10 seconds. The design of this benchmark scheme is as follows Figure 5 shown.

[0119] (1) Visual stimulation form

[0120] The visual stimulus factors that can be further adjusted in the PVT paradigm are as follows:

[0121] 1. Color selection of stimulus presentation background: black and gray.

[0122] 2. Selection of the shape of the fixation point during the stimulus interval: square and cross.

[0123] 3. Color selection of the fixation point during the stimulus interval: red and yellow.

[0124] 4. Choice of stimulus form: dynamic numbers and static circles.

[0125] 5. Stimulating color choices: yellow and red.

[0126] 6. Choice of post-stimulus feedback: yes or no.

[0127] The results of the control variable comparison of the above visual stimulation factors are shown in Table 1.

[0128] Table 1 Results of the effects of visual stimulation factors on mental fatigue induction

[0129]

[0130]

[0131] The results in Table 1 indicate that this universal baseline PVT paradigm already has a good effect on inducing mental fatigue. Regarding visual stimulation, potential improvements include switching the stimulus format from dynamic numbers to simple static graphics and eliminating the immediate feedback process for each trial.

[0132] (2) Test duration

[0133] Table 2 The effect of test duration on the induction of mental fatigue

[0134]

[0135] According to the results in Table 2, it can be found that longer PVT tasks have a better effect in inducing mental fatigue. If experimental conditions permit, longer tests should be given priority to obtain better experimental data.

[0136] (3) Random stimulation interval

[0137] Table 3 Effect of stimulation interval on mental fatigue induction

[0138]

[0139] The results in Table 3 indicate that a shorter randomized interstimulus interval did not lead to a better effect on inducing mental fatigue in the PVT paradigm, but the difference between the two interstimulus interval configurations was small. If there is a need to compress experimental time or perform more tests within the same timeframe, appropriately reducing the randomized interstimulus interval is a viable alternative.

[0140] In summary, by comparing the differences in task performance and EEG characteristics of different PVT design elements, the present invention finds a PVT design scheme with higher mental fatigue induction efficiency, providing a better experimental method for applications such as mental fatigue detection using machine learning.

[0141] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and that in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A method for improving the PVT experimental paradigm based on behavioral and EEG characteristics, characterized by: The method comprises: Changing the visual stimulation form of the PVT experiment to obtain several groups of PVT experimental schemes with different visual stimulation forms, testing subjects using the several groups of PVT experimental schemes with different visual stimulation forms, obtaining several groups of experimental results, analyzing the several groups of experimental results respectively, and selecting the PVT experimental scheme with the best visual stimulation form that induces mental fatigue; Based on the optimal visual stimulation form PVT experimental scheme, the test duration of the PVT experiment is changed to obtain several groups of PVT experimental schemes with different test durations, and the subjects are tested using the several groups of PVT experimental schemes with different test durations to obtain several groups of experimental results. The several groups of experimental results are analyzed separately, and the PVT experimental scheme with the best test duration for inducing mental fatigue is selected; Based on the PVT experimental plan with the optimal test duration, the random stimulation interval of the PVT experiment was changed to obtain several groups of PVT experimental plans with different random stimulation intervals. The subjects were tested using several groups of PVT experimental plans with different random stimulation intervals to obtain several groups of experimental results. These groups of experimental results were analyzed separately, and the PVT experimental plan with the best mental fatigue inducing effect was selected.

2. The improved method of the PVT experimental paradigm based on behavioral and EEG characteristics according to claim 1, characterized in that: The analysis of several groups of experimental results respectively includes: The fatigue trial ratio of several groups of experimental results was analyzed respectively, and the PVT experimental scheme with the best fatigue trial ratio was selected; Conduct fatigue time-frequency domain analysis on several groups of experimental results and select the optimal PVT experimental scheme in fatigue time-frequency domain; Based on the PVT experimental scheme with the best proportion of fatigue trials and the best fatigue time-frequency domain, the PVT experimental scheme with the best visual stimulation form that induces mental fatigue / the PVT experimental scheme with the best test duration that induces mental fatigue / the PVT experimental scheme with the best mental fatigue induction effect are selected.

3. The improved method of the PVT experimental paradigm based on behavioral and EEG characteristics according to claim 2, characterized in that: The fatigue trial ratio analysis is performed on several groups of experimental results, and the PVT experimental scheme with the best fatigue trial ratio is selected, including: Collect behavioral task performance data from each trial during each set of experiments; Clean and filter behavioral task performance data to obtain behavioral data segments and local reaction times; The behavioral data segments and local reaction time were used to calculate the proportion of fatigue trials in each group of experiments, and the PVT experimental plan with the optimal proportion of fatigue trials was selected.

4. The improved method of the PVT experimental paradigm based on behavioral and EEG characteristics according to claim 3 is characterized in that: The fatigue time-frequency domain analysis is performed on several groups of experimental results, and the optimal PVT experimental scheme in the fatigue time-frequency domain is selected, including: Collect the original physiological EEG signals generated in each trial during each experimental session; Preprocessing the physiological original EEG signal to obtain the EEG signal; Conduct frequency domain feature analysis on the EEG signals of each experimental group and select the PVT experimental scheme with the best frequency domain analysis; Perform time domain feature analysis on the EEG signals of each experimental group and select the PVT experimental scheme with the best time domain analysis; The optimal PVT experimental scheme in frequency domain analysis and the optimal PVT experimental scheme in time domain analysis are comprehensively analyzed, and the optimal PVT experimental scheme in fatigue time-frequency domain is selected.

5. The improved method of the PVT experimental paradigm based on behavioral and EEG characteristics according to claim 4 is characterized in that: The calculation of the fatigue trial ratio of each group of experiments using behavioral data segments and local reaction time includes: For each set of experiments, one trial among several trials, as well as 5 trials before and after, were selected, for a total of 11 trials; The local reaction time of 11 trials is obtained, and the local reaction time is expressed as: RT=T R -T S , Where RT represents the local reaction time, T S represents the moment when the visual stimulus is presented on the computer screen, T R The moment when the signal indicating that the subject pressed the designated key was received by the computer; The global reaction time of each of the 11 trials was calculated using the equal weighted average method. The global reaction time is expressed as: GlobalRT i =w i-l RT i-l +…+w i RT i +…+w i+l RT i+l , Where w represents the weight, RT represents the local reaction time, i represents the i-th trial, and l represents the length of the window, that is, the number of adjacent trials included in the calculation; The 5th percentile of the local reaction time of 11 trials was used as the benchmark, and 1.25 times the benchmark was taken as the threshold; The local reaction time and global reaction time of the 11 trials are compared with the threshold value. If the local reaction time and global reaction time of the current trial are both less than the threshold value, the current trial is judged as an alert trial; if the local reaction time and global reaction time of the current trial are both greater than the threshold value, the current trial is judged as a fatigue trial. The proportion of fatigue trials was obtained based on the ratio of fatigue trials to the 11 trials.

6. The improved method of the PVT experimental paradigm based on behavioral and EEG characteristics according to claim 5, characterized in that: The frequency domain feature analysis of the EEG signals of each group of experiments is performed, and the optimal PVT experimental scheme is selected according to the frequency domain analysis, including: The EEG signals of each group of experiments were divided into five frequency bands: δ (1-4 Hz), θ (4-8 Hz), α (8-14 Hz), β (14-31 Hz), and γ (31-50 Hz). For each set of experiments, the average logarithmic power of fatigue trials and alert trials in the five frequency bands was calculated respectively to obtain the average logarithmic power of fatigue trials and the average logarithmic power of alert trials. The average logarithmic power is expressed as: Where N is the number of sampling points, and x[n] is the expression of the discrete signal; The differences between the average logarithmic power of fatigue trials and the average logarithmic power of alert trials in the five frequency bands were calculated respectively, and the average logarithmic power difference of the δ frequency band, the θ frequency band, the α frequency band, the β frequency band, and the γ frequency band were obtained. The average logarithmic power differences of the five frequency bands in each group of experiments were compared, and the experiments with high average logarithmic power differences in the θ band, high average logarithmic power differences in the α band, and low average logarithmic power differences in the β band were selected as the optimal PVT experimental scheme for frequency domain analysis.

7. The improved method of the PVT experimental paradigm based on behavioral and EEG characteristics according to claim 6, characterized in that: The time domain feature analysis of the EEG signals of each group of experiments is performed, and the optimal PVT experimental scheme for the time domain analysis is selected, including: Obtain the OZ channel ERP waveforms of fatigue trials and alertness trials in the occipital lobe of each experimental group; The ERP waveforms of the OZ channel in the fatigue trial and the ERP waveforms of the OZ channel in the alert trial of each group of experiments were compared, and the experiment in which the P1 amplitude of the ERP waveform of the OZ channel in the fatigue trial was smaller than the P1 amplitude of the ERP waveform of the OZ channel in the alert trial was selected as the optimal PVT experimental plan for time domain analysis.

8. The improved method of the PVT experimental paradigm based on behavioral and EEG characteristics according to claim 7 is characterized in that: The preprocessing of the physiological raw EEG signal includes: deleting useless segments, deleting useless electrodes, electrode positioning, low-pass filtering, high-pass filtering, removing power frequency interference, electrode re-reference, reducing the sampling rate, removing obvious bad segments and bad tracks, removing artifacts using ICA technology, band-pass filtering of frequency bands, extracting ERP segments, and performing baseline correction on the ERP segments.

Images