Attention evaluation method and system based on eye movement behavior synchronization and dual-system decoupling modeling

Abstract

An attention assessment method and system based on eye-tracking behavior synchronization and dual-system decoupling modeling is proposed. The method includes: constructing a standardized attention task; simultaneously collecting eye-tracking and behavioral data to construct a synchronized dataset; cleaning, transforming coordinates, and windowing the synchronized dataset to generate cleaned post-trial behavioral data and windowed eye-tracking data; calculating model-free observation indices based on the cleaned post-trial behavioral data and windowed eye-tracking data, and fitting parameters of the passive physiological gaze system; fitting parameters of the active cognitive control system based on the fixed passive physiological gaze system parameters; calculating multi-state information content indices based on the passive physiological gaze system parameters and the active cognitive control system parameters; and assessing attention based on the multi-state information content indices and model-free observation indices. This invention enables the decoupled quantification of passive eye-tracking stability, active cognitive control, distraction interference, and fatigue dynamics within the same assessment framework.

Description

Technical Field

[0001] This invention relates to the field of attention assessment technology, and in particular to an attention assessment method and system based on eye-tracking behavior synchronization and dual-system decoupling modeling. Background Technology

[0002] Attention assessment techniques typically include subjective scales, traditional behavioral tasks, and eye-tracking measurements. Subjective scales rely on subjects or observers rating attention, impulse control, fatigue, emotional state, and sleep status. While convenient, they are susceptible to subjective feelings, recall bias, and social expectations. Traditional continuous execution tasks and Go / No-Go tasks can evaluate sustained attention and reaction inhibition through reaction time, missed reports, false reports, and accuracy. However, these results primarily reflect final behavioral performance and struggle to distinguish whether performance decline stems from insufficient baseline eye movement stability, capture of exogenous distracting stimuli, or exhaustion of active cognitive control resources. Eye-tracking techniques can record fixation point position, pupillary state, and eye movement stability, providing objective physiological indicators of attentional state. However, current applications largely remain at the descriptive statistical level, focusing on average fixation distance, fixation duration, and saccade counts, lacking a technical solution that unifies and separately interprets the physiological processes of eye movement with the behavioral decision-making process.

[0003] Existing technologies combine continuous task execution or visual Go / No-Go paradigms with synchronous eye-tracking data acquisition. Through standardized stimulus presentation, key response recording, and fixation trajectory analysis, they assess subjects' sustained attention, response inhibition, and distractibility. These technologies typically present target and non-target stimuli during experiments, requiring subjects to respond to the target and inhibit responses to the non-target, while simultaneously collecting eye-tracking data. Subsequent processing calculates indicators such as accuracy, reaction time, false alarm rate, central fixation deviation, and fixation changes under distraction conditions. Some protocols also compare behavioral or eye-tracking performance between early and later stages of the task to infer fatigue or changes in attention maintenance ability.

[0004] The most significant shortcoming of the aforementioned technologies lies in their tendency to only statistically analyze behavioral and eye-tracking indicators side-by-side, making it difficult to reveal the underlying mechanisms of attentional abnormalities. Specifically, existing technologies often fail to effectively distinguish between basic fixation instability caused by the passive physiological fixation system, amplified fixational deviations caused by exogenous distraction stimuli, and changes in task rule maintenance, response inhibition, or useful visual field accommodation caused by the active cognitive control system. Furthermore, existing methods handle temporal dynamics coarsely, often using the difference in mean values ​​before and after the task to represent fatigue effects, lacking parameterized models that can continuously characterize the decline in fixational accuracy and changes in active control throughout the task. Therefore, existing technologies struggle to decouple and quantify passive eye-tracking stability, active cognitive control, distraction interference, and fatigue dynamics within the same assessment framework, and also fail to develop interpretable, traceable, and objective technical indicators suitable for individualized attentional profiling or auxiliary assessment of attention deficit problems such as ADHD. Summary of the Invention

[0005] In view of the above-mentioned deficiencies of the prior art, the present invention provides an attention assessment method and system based on eye-movement behavior synchronization and dual-system decoupling modeling, which solves the following technical problems: how to simultaneously acquire eye-movement physiological data and behavioral response data in a standardized attention task, and through an interpretable dual-system model, decompose the attention information acquisition process into a passive physiological gaze system and an active cognitive control system, thereby forming an objective, dynamic, and mechanistic attention assessment method and system.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] The first aspect is an attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling, which includes the following steps:

[0008] S1. Construct a standardized attention task; simultaneously collect eye-tracking data and behavioral data to build a synchronous dataset;

[0009] S2. Clean the synchronous dataset, transform its coordinates, and truncate it to generate cleaned trial behavioral data and windowed eye-tracking data.

[0010] S3. Calculate model-free observation indices based on post-washed trial behavioral data and windowed eye-tracking data, and fit parameters of the passive physiological gaze system.

[0011] S4. Fit the parameters of the active cognitive control system based on the fixed parameters of the passive physiological gaze system;

[0012] S5. Calculate the multi-state information content index based on the parameters of the passive physiological gaze system and the active cognitive control system.

[0013] S6. Attention evaluation is performed based on multi-state information content indicators and model-free observation indicators.

[0014] Preferably, S2 includes:

[0015] The eye-tracking data is cleaned for eye-tracking validity, then interpolated and smoothed to generate a valid gaze point sequence. The valid gaze point sequence is converted into center-view distances by the gaze point pixel coordinates to generate a frame-level center-view distance sequence.

[0016] Frame-level center-view distance sequences under the core stimulus presentation window are extracted to generate windowed eye-tracking data; the trial-subsequent behavioral data are cleaned to generate cleaned trial-subsequent behavioral data.

[0017] Preferably, S3 includes:

[0018] S31. Construct a model-free set of observation indicators based on windowed eye-tracking data and cleaned post-trial behavioral data;

[0019] S32. Establish a basic standard deviation model for the passive physiological fixation system and calculate the basic fixation standard deviation under distraction-free conditions;

[0020] S33. Based on the baseline fixation standard deviation and distraction markers, calculate the passive fixation standard deviation after considering distraction stimuli using a uniform distraction stimulus amplification factor.

[0021] S34. Based on the frame-level center-view distance and passive gaze standard deviation, construct the first-stage maximum likelihood estimate to fit the passive physiological gaze system parameters, including the initial gaze angle standard deviation, passive gaze time fatigue coefficient, and distraction stimulus amplification coefficient.

[0022] As a preferred option, in S32, the baseline fixation standard deviation under the condition of no distraction is calculated using the following formula:

[0023]

[0024] in, Indicates the first The baseline fixation standard deviation under undistracted conditions. This represents the standard deviation of the initial fixation angle. This represents the fatigue coefficient during passive gaze. For windowed eye-tracking data The subjective time variable corresponding to the frame;

[0025] S33 calculates the result after considering distraction stimuli. The standard deviation of passive fixation in a frame is calculated as follows:

[0026]

[0027] in, This represents the standard deviation of passive fixation after considering distracting stimuli. Distraction marker, Indicates the amplification factor of distraction stimuli;

[0028] S34. The first-stage maximum likelihood estimation is constructed as follows:

[0029]

[0030] in, This represents the total number of valid eye-tracking frames. The frame-level center-view distance. This represents the standard deviation of passive fixation.

[0031] Preferably, S4 includes:

[0032] S41. Calculate the active control coefficient based on passive physiological gaze system parameters and subjective time variables;

[0033] S42. Calculate the effective standard deviation after active control based on the basic standard deviation and the active control coefficient.

[0034] S43. Map the effective standard deviation to the probability of obtaining behavioral information based on the effective target radius of the proactive behavior;

[0035] S44. Construct a second-stage maximum likelihood estimate based on the probability of acquiring behavioral information, which is used to fit the parameters of the active cognitive control system, including the active control force coefficient and the active control force time modulation coefficient.

[0036] As a preferred embodiment, in S41, the active control force coefficient is calculated as follows:

[0037]

[0038] in, Indicates the first Trial active control coefficient, This represents the initial active control force coefficient. This represents the time modulation coefficient of the active control force. Subjective time variable;

[0039] S42. The effective standard deviation after active control is calculated as follows:

[0040]

[0041] in, Indicates the first The effective standard deviation after one trial of active control For the first Trial basic standard deviation This is the active control force coefficient;

[0042] S43. The probability of obtaining behavioral information is calculated as follows:

[0043]

[0044] in, Indicates the first Probability of obtaining trial behavior information Indicates the effective target radius of proactive behavior. Represents the error function;

[0045] S44. Construct the second-stage maximum likelihood estimate as follows:

[0046]

[0047] in, The probability of an action being valid and correct. Indicates the number of valid trials.

[0048] As a preferred embodiment, S5 includes:

[0049] S51. Calculate the passive physiological gaze system information based on the passive physiological gaze system parameters and the passive information central visual field threshold, including the passive information in the initial undistracted state, the passive information in the later undistracted state of the task, and the passive information in the initial distracted state.

[0050] S52, Effective target radius based on active cognitive control system parameters and active behavior Calculate the information content of the active cognitive control system, including the information content of active behavior in the undistracted state and the information content of active behavior in the distracted state;

[0051] S53. Both the passive information and the active behavioral information in the initial distracted state are amplified using the distraction stimulus amplification factor. Generate a set of information content indicators with homologous distraction interpretation.

[0052] As a preferred option, in S51, the amount of passive information in the initial undistracted state is calculated as follows:

[0053]

[0054] in, The passive information center field threshold, The initial fixation angle standard deviation is given; the passive information content in the later stages of the task under undistracted conditions is calculated as follows:

[0055]

[0056] in, This represents the fatigue coefficient during passive gaze. This represents the objective time corresponding to the later stage of the task. The amount of passive information in the initial distracted state is calculated as follows:

[0057]

[0058] in, This is the amplification factor for distraction stimuli;

[0059] In S52, the amount of information about active behavior in the undistracted state is calculated as follows:

[0060]

[0061] in, Indicates the first The amount of information about proactive behavior in a state of undistracted engagement. The effective standard deviation after active control. Given the effective target radius of the active behavior, the information content of the active behavior in the distracted state is calculated as follows:

[0062]

[0063] in, Indicates the first Information content of proactive behavior in a state of distraction.

[0064] As a preferred embodiment, S6 includes:

[0065] An attention assessment feature set is generated based on model-free observation indicators, passive physiological fixation system parameters, active cognitive control system parameters, and multi-state information content indicators. Based on the attention assessment feature set, the basic fixation stability, passive anti-fatigue stability, extrinsic distractibility, active cognitive control ability, and behavioral output efficiency are evaluated to generate an individual attention profile. Based on the individual attention profile, attention assessment conclusions are generated, and then experimental verification results are generated.

[0066] Secondly, an attention assessment system based on eye-tracking behavior synchronization and dual-system decoupling modeling includes:

[0067] A standardized task presentation module is used to build standardized attention tasks;

[0068] The eye-tracking and behavior synchronization acquisition module is used to simultaneously acquire eye-tracking data and behavior data to build a synchronized dataset;

[0069] The eye-tracking behavior preprocessing module is used to clean, transform coordinates, and truncate windows of the synchronous dataset to generate cleaned trial behavior data and windowed eye-tracking data.

[0070] The passive gaze system fitting module is used to calculate model-free observation indices based on washed post-trial behavioral data and windowed eye movement data, and to fit the parameters of the passive physiological gaze system.

[0071] The active cognitive control fitting module is used to fit the parameters of the active cognitive control system based on the parameters of the passive physiological gaze system.

[0072] The indicator fusion and information calculation module is used to calculate multi-state information indicators based on parameters of the passive physiological gaze system and parameters of the active cognitive control system.

[0073] The evaluation report output module is used to evaluate attention based on multi-state information content indicators and model-free observation indicators;

[0074] The attention assessment system based on eye-tracking behavior synchronization and dual-system decoupling modeling is used to implement the attention assessment method and steps based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in the first aspect.

[0075] Compared with the prior art, the beneficial effects of the present invention are reflected in:

[0076] 1. Unlike traditional techniques that merely statistically analyze behavioral task results alongside descriptive eye-tracking indicators, making it difficult to distinguish the sources of attentional decline, this invention employs a standardized CPT / Go-NoGo behavioral paradigm, simultaneous eye-tracking acquisition, and a decoupled modeling approach combining the passive physiological gaze system and the active cognitive control system. This allows for the simultaneous acquisition of mechanistic indicators such as basic fixation stability, distractibility sensitivity, fatigue-related gaze accuracy changes, active control constraint ability, and behavioral information extraction efficiency in a single attention test. This elevates the traditional superficial judgment of "whether attention is insufficient" to an interpretable, quantifiable assessment of "which attentional subsystem is abnormal." The effectiveness of this technique is not a simple superposition of behavioral tasks, eye-tracking acquisition, or statistical analysis. Instead, through unified experimental synchronization, hierarchical parameter estimation, and dual-system information mapping, it creates mutual constraints and verification between the eye-tracking physiological process and the behavioral decision-making process within the same model framework, resulting in a synergistic effect of objectively separating and dynamically explaining the attention mechanism.

[0077] 2. Unlike traditional techniques that evaluate the effects of attention fatigue and distraction solely based on differences in mean values ​​between different time periods or overall accuracy and average reaction time, this invention employs a joint modeling approach based on task time progression, distraction stimulus conditions, and behavioral correctness. This allows attention assessment to continuously depict the decay trend of passive gaze accuracy as the task progresses, and further analyzes the combined effects of exogenous distraction stimuli on the passive gaze system and the active cognitive control system, thereby improving the accuracy of identifying sustained attention maintenance ability, anti-interference ability, and post-fatigue cognitive compensation ability. The innovation of this approach lies in not simply adding distraction stimuli or increasing before-and-after comparisons, but rather organizing time decay parameters, distraction amplification parameters, and active control parameters into a combined structure with causal explanatory relationships. This ensures that distraction, fatigue, and active control are no longer isolated indicators, but rather jointly determine information acquisition efficiency within the same technical chain, resulting in a comprehensive technical effect more suitable for individualized attention profiling, attention deficit screening, and subsequent product evaluation. Attached Figure Description

[0078] Figure 1 This is a schematic diagram of the method framework of Example 1;

[0079] Figure 2 This is a schematic diagram of the system structure of Example 1. Detailed Implementation

[0080] To make the technical means, inventive features, objectives, and effects of the invention readily understandable, the invention is further described below with reference to specific illustrations. However, the invention is not limited to the embodiments described below.

[0081] It should be noted that the structures, proportions, sizes, etc., illustrated in the accompanying drawings of this specification are only used to complement the content disclosed in the specification for those skilled in the art to understand and read, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0082] The technical problem this invention aims to solve is: how to simultaneously acquire eye-tracking physiological data and behavioral response data in a standardized attention task, and how to decompose the attention information acquisition process into a passive physiological gaze system and an active cognitive control system through an interpretable dual-system model, thereby forming an objective, dynamic, and mechanistic attention assessment method and system. Specifically, it includes the following technical issues:

[0083] (1) How to design a standardized task that combines sustained attention, target detection, response inhibition and anti-interference load, so that it can induce behavioral differences and stably collect eye movement fixation data within the stimulus presentation window.

[0084] (2) How to perform validity labeling and cleaning, screen boundary filtering, blink artifact removal, short-term missing interpolation and smoothing on eye movement data so that subsequent fixation distance and viewpoint conversion indicators are comparable.

[0085] (3) How to estimate the basic fixation distraction, continuous task time fatigue and extrinsic distraction amplification effect in frame-scale eye-tracking data so that the passive physiological fixation system can be represented by independent parameters.

[0086] (4) How to estimate the active cognitive control parameters based on the accuracy of trial behavior, reaction time cost and task status on the basis of fixed passive system parameters, so that the ability of active control to constrain the effective field of vision can be quantified.

[0087] (5) How to unify the model parameters and original model-free indicators into comparable information content indicators and capability vectors for individual reporting, population typing, subsequent ADHD-related clinical validation and product application.

[0088] This invention provides an attention assessment method and system based on eye-tracking behavior synchronization and dual-system decoupling modeling. The scheme simultaneously collects eye-tracking fixation data and key-pressing behavior data through a standardized CPT and visual Go / No-Go fusion task. After preprocessing, model-free observation metrics are calculated first, and then two-step maximum likelihood estimation is used to fit the passive physiological fixation system and the active cognitive control system, respectively. Finally, model parameters, information content metrics, capability vectors, and user profiles are output.

[0089] This approach is particularly suitable for objective assessment of attention, research on the mechanisms of attention deficit in ADHD, construction of norms for normal samples, pre- and post-intervention follow-up, and generation of productized reports. It is important to emphasize that the output of this approach can be used as supplementary assessment and mechanistic analysis data, but should not be considered as a standalone diagnosis of the disease.

[0090] The core idea of ​​this invention is that performance on an attention task is not a single behavioral outcome, but rather determined by at least two separable processes. The first is a bottom-up passive physiological fixation system, primarily reflecting baseline fixation stability, distraction stimulus capture, and sustained task fatigue. The second is a top-down active cognitive control system, primarily reflecting target detection, response inhibition, effective visual field contraction, and behavioral decision-making ability. This invention achieves relative decoupling of these two systems through hierarchical fitting of eye-tracking frame-level data and behavioral trial-level data.

[0091] Example 1:

[0092] like Figure 1 The attention assessment method shown includes the following steps: (The method is based on eye-tracking behavior synchronization and dual-system decoupling modeling.)

[0093] Step S1: Construct a standardized attention task and simultaneously collect eye-tracking and behavioral data.

[0094] Step S11: Set attention task parameters. Set task configuration parameters, including total number of trials, single trial duration, stimulus presentation duration, stimulus interval, target stimulus ratio, non-target stimulus ratio, probability of distracting stimulus appearance, screen resolution, eye-tracking sampling rate, and response button. As a specific implementation, the total number of trials is 300, the single trial duration is 1200ms, the core stimulus presentation window is 700ms, and the stimulus interval is 500ms; the target stimulus is a Go stimulus, the non-target stimulus is a No-Go stimulus, and distracting stimuli are presented synchronously with the core stimulus at a preset probability. Generate a standardized task sequence based on the task configuration parameters.

[0095] Step S12: Perform the CPT and Go / No-Go fusion task. Based on the standardized task sequence obtained in Step S11, in each trial, the subject continuously fixates on the central fixation point on the screen; when the target stimulus is presented, the subject presses a response key within the core stimulus presentation window; when a non-target stimulus is presented, the subject inhibits the key press response. The final result is trial-level behavioral data including the stimulus type, presence of distracting stimuli, stimulus start time, key press state, reaction time, and trial number for each trial.

[0096] Step S13: Synchronously acquire eye-tracking data. Eye-tracking data includes at least validity markers for the left eye, right eye, or both eyes, pixel coordinates of the fixation point, pupil diameter, and eye-screen distance. A trigger signal is used to synchronize the behavioral data with the eye-tracking data in time, generating a synchronized dataset of frame-level eye-tracking data and trial-level behavioral data.

[0097] Step S2: Clean the synchronized dataset, perform window truncation, and coordinate transformation.

[0098] Step S21: Perform eye movement validity cleaning. Frames in the frame-level eye movement data whose validity markers do not meet preset conditions, frames with gaze points outside the screen boundary, frames with pupil diameters lower than the blink threshold, and frames within a preset extended window before and after the blink are invalidated. Preliminary cleaned eye movement data is generated.

[0099] Step S22: Perform fixation point interpolation and smoothing. Linear interpolation is performed on short-term missing fixations in the pre-cleaned eye-tracking data, and missing intervals exceeding the maximum continuous interpolation length are kept null. Subsequently, a smoothing filter is used to reduce coordinate noise and restore the original long missing positions to null values, generating a smoothed effective fixation point sequence.

[0100] Step S23: Convert the gaze point pixel coordinates to the center view distance. For a valid gaze point sequence, use the screen center coordinates. Using the origin as the starting point, the first... Frame gaze point pixel coordinates Converted to viewing distance The calculation formula is:

[0101]

[0102] in, Indicates the first The visual distance from the frame's gaze point to the center of the screen. This represents the number of viewing angles corresponding to a single pixel. A frame-level center-view distance sequence is generated.

[0103] Step S24: Extract the core stimulus presentation window and clean the behavioral data. Using the trial number, core stimulus start time, and synchronization trigger marker recorded in step S12 as alignment indices, divide the frame-level center-view distance sequence generated in step S23 into the corresponding trials, and calculate the time offset of each frame relative to the core stimulus start time of that trial. Only retain the center-view distance within the core stimulus presentation window, its timestamp, distraction marker, and stimulus type to generate windowed eye-tracking data; trials without valid eye-tracking frames are marked as invalid eye-tracking data. The frame-level center-view distance sequence is used to form the eye-tracking observation sequence for each trial and is not directly used as the basis for determining the validity of the behavioral response. For behavioral data, the reaction time is calculated based on the difference between the button press time and the core stimulus start time; if there are multiple button presses in the same trial, the first valid button press within the core stimulus presentation window is taken as the response for that trial. Key presses with a reaction time shorter than the preset lower limit are considered too fast reactions, while key presses exceeding the core stimulus presentation window are considered timed out reactions. The corresponding reaction times are then set to null and excluded from the valid reaction time statistics. In one specific implementation, the preset lower limit for reaction time is 200 ms, and the core stimulus presentation window is 700 ms. For valid trials, in Go trials, key presses within the valid reaction time range are recorded as hits, and no valid key presses are recorded as misses. In No-Go trials, key presses within the valid reaction time range are recorded as false alarms, and no valid key presses are recorded as correct rejections. Based on this, trial validity markers, cleaned reaction times, and behavioral correctness markers are generated, forming cleaned trial-level behavioral data.

[0104] Step S3: Calculate the model-free observation index and fit the parameters of the passive physiological gaze system.

[0105] Step S31: Calculate model-free observation metrics. Based on the windowed eye-tracking data and cleaned trial-level behavioral data output in step S24, calculate the average fixation distance at the beginning of the task, the average fixation distance at the end of the task, the average fixation distance under initial distraction conditions, the average fixation distance under later distraction conditions, the average accuracy, the initial effective reaction time cost, and the later effective reaction time cost, generating a set of model-free observation metrics.

[0106] Step S32: Establish the basic standard deviation model of the passive physiological fixation system. This is based on the frame-level center-view distance in windowed eye-tracking data. , No. The subjective time variable corresponding to the frame and the Whether the frame is in a distraction trial flag .in, , Indicates the first The objective time corresponding to the frame The baseline fixation standard deviation under distraction-free conditions is expressed as:

[0107]

[0108] in, Indicates the first The baseline fixation standard deviation under undistracted conditions. This represents the standard deviation of the initial fixation angle. This represents the passive fixation time fatigue coefficient and generates a baseline fixation standard deviation sequence.

[0109] Step S33: Introduce a unified distraction stimulus amplification factor. Based on the baseline fixation standard deviation sequence and distraction markers obtained in step S32. After considering distracting stimuli, the first... The standard deviation of passive gaze in a frame is expressed as:

[0110]

[0111] in, This represents the standard deviation of passive fixation after considering distracting stimuli. This represents the distraction stimulus amplification factor, generating a passive fixation standard deviation sequence.

[0112] Step S34: Fit the parameters of the passive physiological gaze system using the first-stage maximum likelihood estimation. Based on the center-view distance. and passive fixation standard deviation The objective function is to minimize the negative log-likelihood.

[0113]

[0114] in, This represents the total number of valid eye-tracking frames. This represents the negative log-likelihood of the first stage. The fitted parameters of the passive physiological gaze system are then generated. , and .

[0115] Step S4: Fit the parameters of the active cognitive control system based on the fixed parameters of the passive physiological gaze system.

[0116] Step S41: Establish an active control force time modulation model. The passive physiological gaze system parameters obtained in step S34 are fixed and will not be re-estimated in the second stage; wherein, the standard deviation of the initial gaze angle is used. Passive gaze duration fatigue coefficient ,according to Calculate the baseline passive fixation standard deviation for each effective eye-tracking frame, and then... Within the core stimulus presentation window of the trial The baseline standard deviation of the trial was obtained by aggregating the effective eye-tracking frames. The passive baseline used in step S42 to calculate the effective standard deviation after active control; the distraction stimulus amplification factor obtained in step S34. Keep this constant; it will be used for calculating the amount of information about active behavior in subsequent distracted states and will not be repeatedly fitted in this step. Based on the first... Subjective time variable of trials Construct the active control coefficient for this trial. ,in, This represents the initial active control force coefficient. This represents the active control force time modulation coefficient, which ultimately generates an active control force sequence that corresponds one-to-one with the basic standard deviation of each trial.

[0117] Step S42: Calculate the effective standard deviation after active control. Based on the... Trial basic standard deviation and active control coefficient The effective standard deviation after active control is calculated as follows:

[0118]

[0119] in, Indicates the first The effective standard deviation after one trial of active control.

[0120] Step S43: Map the effective standard deviation to the probability of obtaining behavioral information. Based on the effective standard deviation output in step S42... and the effective target radius of proactive behavior Calculate the first The probability of obtaining the target information in one trial is:

[0121]

[0122] in, Indicates the first Probability of obtaining trial behavior information Indicates the effective target radius of proactive behavior. The error function represents the probability sequence of obtaining trial behavior information.

[0123] Step S44: Based on the effective correctness probability of the behavior and the probability of obtaining behavioral information A second-stage maximum likelihood estimation is constructed to fit the parameters of the active cognitive control system. Among these, The negative log-likelihood of the second stage can be determined by the number of hits in the Go trials, the number of missed detections in the Go trials, the number of correct suppressions in the No-Go trials, and the number of false alarms in the No-Go trials.

[0124]

[0125] in, Indicates the number of valid trials. This represents the negative log-likelihood of the second stage. Finally, the fitted parameters of the active cognitive control system are generated. and active control force time modulation coefficient .

[0126] Step S5: Calculate the multi-state information content index based on the parameters of the passive physiological gaze system and the active cognitive control system.

[0127] Step S51: Calculate the information content of the passive physiological gaze system. Based on the initial gaze angle standard deviation obtained in step S34. Passive gaze duration fatigue coefficient Distraction stimulus amplification factor and passive information center field of view threshold The amount of passive information in the initial undistracted state is calculated as follows:

[0128]

[0129] The amount of passive information during the later stages of a mission when the user is not distracted is as follows:

[0130]

[0131] The amount of passive information in the initial state of distraction is:

[0132]

[0133] in, This indicates the objective time corresponding to the later stages of the task. This represents the passive information center field threshold.

[0134] Step S52: Calculate the information content of the active cognitive control system. Based on the effective standard deviation obtained in step S42 after active control. The distraction stimulus amplification factor obtained in step S34 and the effective target radius of proactive behavior The information content of active behavior in a distraction-free state is calculated as follows:

[0135]

[0136] The information content of proactive behavior in a distracted state is:

[0137]

[0138] in, Indicates the first The amount of information about proactive behavior in a state of undistracted engagement. Indicates the first Information content of proactive behavior in a state of distraction.

[0139] Step S53: Unify the interpretation of distraction parameters. The same distraction stimulus amplification factor is used for both passive and active distraction information. Without setting additional independent active distraction free parameters, a set of information content indicators with homologous distraction explanations is generated, including information content of the passive physiological gaze system and information content of the active cognitive control system.

[0140] Step S6: Generate attention assessment results.

[0141] Step S61: Integrate model parameters, information content indicators, and model-free observation indicators. Based on the model-free observation indicators output in step S31, the passive physiological gaze system parameters output in step S34, the active cognitive control system parameters output in step S44, and the information content indicators output in step S5, generate an attention assessment feature set.

[0142] Step S62: Generate individual attention profiles. Based on the attention evaluation feature set obtained in step S61. , , , , The passive information content, active behavioral information content, average accuracy, and reaction time cost are used to evaluate basic fixation stability, passive anti-fatigue stability, extrinsic distraction sensitivity, active cognitive control ability, and behavioral output efficiency, respectively, to generate an individual attention profile.

[0143] Step S63: Output the evaluation results. Based on the individual attention profile obtained in step S62, the output includes at least passive physiological gaze system parameters, active cognitive control system parameters, multi-state information content indicators, model-free observation indicators, and attention evaluation conclusions based on the above indicators. These evaluation conclusions characterize the subject's efficiency in acquiring attentional information and its dynamic changes under distracting stimuli and sustained task duration.

[0144] Step S64: Generate experimental validation results. Based on the attention assessment feature set obtained in Step S61 and the individual attention profile obtained in Step S62, experimental validation is conducted using standardized attention task data from 83 subjects. The data includes frame-level eye-tracking data, trial-level behavioral data, passive physiological gaze system parameters, active cognitive control system parameters, and multi-state information content indicators. Experimental validation results are then generated.

[0145] In one specific implementation, the core experimental results are shown in Table 1 below:

[0146] Table 1:

[0147]

[0148] Furthermore, this embodiment also outputs parameters of the passive physiological gaze system and the active cognitive control system, and the core statistical results are shown in Table 2 below:

[0149] Table 2:

[0150]

[0151] The experimental results above demonstrate that, despite a generally high average behavioral accuracy and limited discriminative power, this method can still further separate basic fixation stability, passive fatigue time effect, extrinsic distraction sensitivity, and active cognitive control ability through simultaneous eye-tracking behavioral acquisition and dual-system decoupling modeling. Furthermore, after constructing ability profiles based on parameters and information content indicators, the 83 subjects can be divided into three categories: 29 subjects with stable fixation and insufficient response output; 25 subjects with active compensation and superior response output / insufficient fatigue resistance; and 29 subjects with weak fixation and compensatory anti-interference and anti-fatigue resistance. This indicates that this method can generate individualized attention assessment results with mechanistic explanation, in addition to traditional average accuracy and average reaction time.

[0152] Example 2:

[0153] The system includes modules for task presentation, synchronous data acquisition, data preprocessing, calculation of model-free observation indicators, passive system fitting, active system fitting, information content calculation, capability vector classification, and report output. These modules are connected sequentially according to the task data flow, with the passive system fitting results serving as fixed priors or constraint inputs for the active system fitting.

[0154] like Figure 2 The attention assessment system shown includes eye-tracking behavior synchronization and dual-system decoupling modeling.

[0155] (1) Standardized task presentation module

[0156] This module is used to present an attention task that integrates CPT and the visual Go / No-Go paradigm. The task includes target stimuli, non-target stimuli, and exogenous distraction stimuli. Target stimuli require the subject to press a key, non-target stimuli require the subject to inhibit key presses, and distraction stimuli appear simultaneously with the core stimulus but do not require a response. This module is used to simultaneously induce sustained attention, target detection, response inhibition, and anti-interference processes.

[0157] In one feasible implementation, the task consists of 300 trials with a total duration of approximately 8 minutes; each trial lasts 1200 ms, with core stimuli presented for 700 ms and stimulus intervals of 500 ms; Go-target stimuli account for approximately 20%, No-Go non-target stimuli account for approximately 80%, and distracting stimuli appear synchronously in approximately 30% of the trials.

[0158] (2) Eye-tracking and behavior synchronization acquisition module

[0159] This module is used to synchronously acquire eye-tracking frame data and behavioral trial data. Eye-tracking data includes left and right eye and binocular validity markers, fixation point pixel coordinates, pupil diameter, and eye-screen distance; behavioral data includes trial number, stimulus type, distraction status, key press time, stimulus start time, and trial state. The system aligns the eye-tracking data with the behavioral data via trigger signals, ensuring that eye-tracking sampling points within each stimulus presentation window are mapped to the corresponding trial.

[0160] In one feasible implementation, the eye-tracking sampling rate is 200 Hz, the screen resolution is 1920 x 1080, the task background is neutral gray, and the fixation point is set in the center of the screen; multi-point calibration is performed before the experiment, and the calibration error is required to be lower than a preset threshold.

[0161] (3) Eye-tracking behavior preprocessing module

[0162] This module is used to clean eye movement and behavioral data. Eye movement cleaning includes: removing invalid frames based on validity markers; filtering gaze points that exceed the screen boundaries; identifying blinking or eyelid occlusion states based on pupil diameter; extending blinking or missing segments by a preset time; removing non-task gazes that are off-center and exceed a preset visual angle threshold; performing linear interpolation on short-term missing segments; using a smoothing filtering algorithm to reduce noise; and restoring the original long missing segments after smoothing to avoid fabricated valid data.

[0163] Behavioral cleansing includes: identifying overly rapid responses as guessing responses, identifying responses that exceed the stimulus presentation window as timed responses, and classifying responses as hits, misses, correct rejections, and false alarms based on target / non-target states.

[0164] (4) Model-free observation index calculation module

[0165] This module calculates raw observation metrics independent of maximum likelihood estimation before model fitting. Eye-tracking metrics include average fixation distance at the beginning of the task, average fixation distance at the end of the task, average fixation distance under initial distraction conditions, and average fixation distance under later distraction conditions. Behavioral metrics include average accuracy, initial reaction time cost, and later reaction time cost. This module provides raw references for validating model validity.

[0166] (5) Passive gaze system fitting module

[0167] This module fits passive physiological fixation system parameters onto frame-scale eye-tracking data. Its core idea is to assume that the distance of the fixation point from the screen center can be described by an approximately normal distribution, and that the standard deviation of this distribution is modulated by both task time and distraction stimuli. Passive system parameters include the initial fixation angle standard deviation, the time fatigue factor, and the distraction stimulus amplification factor.

[0168] (6) Active cognitive control fitting module

[0169] After fixing the passive system parameters, this module fits the active cognitive control parameters using trial-based behavioral results. The active control parameters describe the ability of top-down cognitive control to constrain the effective field of vision. Stronger active control results in a more concentrated range of effective information extraction, making it easier for target detection and response inhibition to lead to correct behavioral outcomes.

[0170] (7) Indicator fusion and information calculation module

[0171] This module converts passive system parameters, active system parameters, and task states into passive information and active behavioral information. Information is represented by a probabilistic index, ranging from 0 to 1; higher values ​​indicate higher efficiency in extracting attentional information. This module simultaneously calculates information at the initial, later, initial distraction, and later distraction stages.

[0172] (8) Capability Vector and Fractal Module

[0173] This module filters and removes redundant representative variables from model parameters, information content indicators, and model-free indicators to construct an individual ability vector space. It can employ principal component analysis, rotational interpretation, and clustering methods to output multiple independent ability dimensions and user profiles, such as passive fixation stability, task execution efficiency, extrinsic distractibility sensitivity, and fatigue time adaptation.

Claims

1. An attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling, characterized in that, Includes the following steps: S1. Construct a standardized attention task; Simultaneously collect eye-tracking and behavioral data to construct a synchronous dataset; S2. Clean the synchronous dataset, transform its coordinates, and truncate it to generate cleaned trial behavioral data and windowed eye-tracking data. S3. Calculate model-free observation indices based on post-washed trial behavioral data and windowed eye-tracking data, and fit parameters of the passive physiological gaze system. S4. Fit the parameters of the active cognitive control system based on the fixed parameters of the passive physiological gaze system; S5. Calculate the multi-state information content index based on the parameters of the passive physiological gaze system and the active cognitive control system. S6. Attention evaluation is performed based on multi-state information content indicators and model-free observation indicators.

2. The attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in claim 1, characterized in that, S2 include: The eye-tracking data is cleaned for eye-tracking validity, and then fixation point interpolation and smoothing are performed to generate a valid fixation point sequence. Convert the effective gaze point sequence to gaze point pixel coordinates into center-view distances to generate a frame-level center-view distance sequence. Frame-level center-view distance sequences under the core stimulus presentation window are extracted to generate windowed eye-tracking data; the trial-subsequent behavioral data are cleaned to generate cleaned trial-subsequent behavioral data.

3. The attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in claim 1, characterized in that, S3 include: S31. Construct a model-free set of observation indicators based on windowed eye-tracking data and cleaned post-trial behavioral data; S32. Establish a basic standard deviation model for the passive physiological fixation system and calculate the basic fixation standard deviation under distraction-free conditions; S33. Based on the baseline fixation standard deviation and distraction markers, calculate the passive fixation standard deviation after considering distraction stimuli using a uniform distraction stimulus amplification factor. S34. Based on the frame-level center-view distance and passive gaze standard deviation, construct the first-stage maximum likelihood estimate to fit the passive physiological gaze system parameters, including the initial gaze angle standard deviation, passive gaze time fatigue coefficient, and distraction stimulus amplification coefficient.

4. The attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in claim 3, characterized in that, In S32, the baseline fixation standard deviation under distraction-free conditions is calculated using the following formula: in, Indicates the first The baseline fixation standard deviation under undistracted conditions. This represents the standard deviation of the initial fixation angle. This represents the fatigue coefficient during passive gaze. For windowed eye-tracking data The subjective time variable corresponding to the frame; S33 calculates the result after considering distraction stimuli. The standard deviation of passive fixation in a frame is calculated as follows: in, This represents the standard deviation of passive fixation after considering distracting stimuli. Distraction markers Indicates the amplification factor of distraction stimuli; The first-stage maximum likelihood estimation in S34 is constructed as follows: in, This represents the total number of valid eye-tracking frames. The frame-level center-view distance. This represents the standard deviation of passive fixation.

5. The attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in claim 1, characterized in that, S4 include: S41. Calculate the active control coefficient based on passive physiological gaze system parameters and subjective time variables; S42. Calculate the effective standard deviation after active control based on the basic standard deviation and the active control coefficient. S43. Map the effective standard deviation to the probability of obtaining behavioral information based on the effective target radius of the proactive behavior; S44. Construct a second-stage maximum likelihood estimate based on the probability of acquiring behavioral information, which is used to fit the parameters of the active cognitive control system, including the active control force coefficient and the active control force time modulation coefficient.

6. The attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in claim 5, characterized in that, In S41, the active control force coefficient is calculated as follows: in, Indicates the first Trial active control coefficient, This represents the initial active control force coefficient. This represents the time modulation coefficient of the active control force. Subjective time variable; The effective standard deviation after active control is calculated in S42 as follows: in, Indicates the first The effective standard deviation after one trial of active control For the first Trial basic standard deviation This is the active control force coefficient; The probability of obtaining behavioral information in S43 is calculated as follows: in, Indicates the first Probability of obtaining trial behavior information Indicates the effective target radius of proactive behavior. Represents the error function; The second-stage maximum likelihood estimate constructed in S44 is as follows: in, The probability of an action being valid and correct. Indicates the number of valid trials.

7. The attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in claim 1, characterized in that, S5 include: S51. Calculate the passive physiological gaze system information based on the passive physiological gaze system parameters and the passive information central visual field threshold, including the passive information in the initial undistracted state, the passive information in the later undistracted state of the task, and the passive information in the initial distracted state. S52, Effective target radius based on active cognitive control system parameters and active behavior Calculate the information content of the active cognitive control system, including the information content of active behavior in the undistracted state and the information content of active behavior in the distracted state; S53. Both the passive information and the active behavioral information in the initial distracted state are amplified using the distraction stimulus amplification factor. Generate a set of information content indicators with homologous distraction interpretation.

8. The attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in claim 7, characterized in that, In S51, the amount of passive information in the initial undistracted state is calculated as follows: in, The passive information center field threshold, The initial fixation angle standard deviation is given; the passive information content in the later stages of the task under undistracted conditions is calculated as follows: in, This represents the fatigue coefficient during passive gaze. This represents the objective time corresponding to the later stage of the task. The amount of passive information in the initial distracted state is calculated as follows: in, This is the amplification factor for distraction stimuli; In S52, the amount of information about active behavior in the undistracted state is calculated as follows: in, Indicates the first The amount of information about proactive behavior in a state of undistracted engagement. The effective standard deviation after active control. Given the effective target radius of the active behavior, the information content of the active behavior in the distracted state is calculated as follows: in, Indicates the first Information content of proactive behavior in a state of distraction.

9. The attention assessment method based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in claim 1, characterized in that, S6 include: An attention assessment feature set is generated based on model-free observation indicators, passive physiological gaze system parameters, active cognitive control system parameters, and multi-state information content indicators. Based on the attention assessment feature set, the basic fixation stability, passive anti-fatigue stability, extrinsic distraction sensitivity, active cognitive control ability and behavioral output efficiency are evaluated to generate an individual attention profile; Attention assessment conclusions are generated based on individual attention profiles, and then experimental verification results are generated.

10. An attention assessment system based on eye-tracking behavior synchronization and dual-system decoupling modeling, characterized in that, include: A standardized task presentation module is used to build standardized attention tasks; The eye-tracking and behavior synchronization acquisition module is used to simultaneously acquire eye-tracking data and behavior data to build a synchronized dataset; The eye-tracking behavior preprocessing module is used to clean, transform coordinates, and truncate windows of the synchronous dataset to generate cleaned trial behavior data and windowed eye-tracking data. The passive gaze system fitting module is used to calculate model-free observation indices based on washed post-trial behavioral data and windowed eye movement data, and to fit the parameters of the passive physiological gaze system. The active cognitive control fitting module is used to fit the parameters of the active cognitive control system based on the parameters of the passive physiological gaze system. The indicator fusion and information calculation module is used to calculate multi-state information indicators based on parameters of the passive physiological gaze system and parameters of the active cognitive control system. The evaluation report output module is used to evaluate attention based on multi-state information content indicators and model-free observation indicators; The attention assessment system based on eye-tracking behavior synchronization and dual-system decoupling modeling is used to implement the attention assessment method and steps based on eye-tracking behavior synchronization and dual-system decoupling modeling as described in claim 1.

Images