A concentration training method and device based on visual tracking

Abstract

The application provides a concentration training method and device based on visual tracking, which comprises the following steps: collecting real-time eye movement data of a user, including a gaze point coordinate and an eye movement speed, and initializing a target trajectory of a training task; calculating a concentration deviation between a current gaze point and the target trajectory, dynamically adjusting the target trajectory through a nonlinear mapping based on the deviation, and generating a dynamic trajectory; based on the concentration deviation and its change trend, combining the geometric features of the trajectory, constructing a distraction score model with multi-feature fusion, and detecting and marking a distraction event; when the distraction event is detected, immediately performing adaptive visual correction stimulation at the event position to guide the user to return to the task; after the training period ends, dynamically adjusting the task difficulty coefficient of the next period according to the frequency and average deviation of the distraction event in the current period.

Description

Technical Field

[0001] This invention belongs to the field of visual training, and particularly relates to a method and apparatus for attention training based on visual tracking. Background Technology

[0002] Attention is one of the key psychological abilities for humans to maintain high performance in various activities such as learning, work, and sports. Its level directly affects the speed and accuracy of information processing and the quality of task completion. In recent years, with the development of eye-tracking technology, image display technology, and real-time computer processing capabilities, attention training methods based on visual tracking have gradually become a research and application hotspot. These methods typically present moving targets on a screen or head-mounted display device, combined with gaze point data collected by eye-tracking devices, to assess and guide the user's gaze behavior, thereby improving attention. However, existing technologies still have significant shortcomings. First, task trajectories are mostly fixed or simple patterns, unable to dynamically adjust the trajectory shape and difficulty based on the user's real-time attention state, leading to insufficient or excessive training stimulation and reduced effectiveness. Second, distraction detection mostly relies on a single deviation threshold, ignoring the combined effect of deviation change trends and target trajectory geometric features, easily resulting in misjudgments and omissions, and failing to accurately identify distractions in the early stages. Third, correction mechanisms are mostly delayed feedback, lacking targeted stimulation at the moment of the distraction event, and are not combined with periodic difficulty adjustments, making it difficult to form a continuous and effective training loop. Therefore, there is an urgent need for a technical solution that can generate adaptive dynamic trajectories in real time during training, accurately detect distraction by fusing multi-dimensional features, execute corrective stimuli immediately when events are triggered, and automatically optimize the task difficulty after the cycle ends, so as to improve the continuity, personalization and long-term effectiveness of training and overcome the shortcomings of existing technologies in terms of real-time performance, detection accuracy and adaptability. Summary of the Invention

[0003] The purpose of this invention is to design a visual tracking-based attention training method and device that organically integrates four aspects: trajectory adaptation, accurate detection, real-time correction, and personalized difficulty adjustment, forming a dual closed loop of data and control. This improves the real-time performance and accuracy of attention training, and also enhances the adaptability and quantifiable effectiveness of long-term training.

[0004] To achieve the above objectives, a first aspect of the present invention provides a focus training method based on visual tracking, the method comprising:

[0005] Collect real-time eye movement data of the user, including gaze point coordinates and eye movement velocity, initialize the target trajectory of the training task, and calculate the focus deviation between the current gaze point and the target trajectory;

[0006] Based on the aforementioned focus deviation, the target trajectory is adjusted through nonlinear mapping to generate a dynamic trajectory;

[0007] Based on focus deviation and its changing trends, and combined with the geometric features of dynamic trajectories, a multi-feature fusion distraction scoring model is constructed for the detection and labeling of distraction events.

[0008] When a distraction event is detected, an adaptive visual correction stimulus is immediately executed at the location of the distraction event to guide the user back to the task. After the training cycle ends, the task difficulty coefficient for the next cycle is dynamically adjusted based on the frequency and average deviation of distraction events in this cycle.

[0009] Furthermore, the steps for adjusting the target trajectory include: calculating the trajectory adjustment magnitude based on the focus deviation and limiting the upper limit of the adjustment amount through a saturation control function; applying the adjustment magnitude along the direction of the gaze point relative to the target point to generate a continuously updated dynamic trajectory.

[0010] Furthermore, a focus-fluctuation adjustment term and a stationarity constraint term are introduced during the dynamic trajectory generation process, so that the trajectory changes can simultaneously take into account both stimuli and controllability.

[0011] Furthermore, the nonlinear mapping employs a saturated nonlinear function to ensure sensitivity with small deviations and controllability with large deviations.

[0012] Furthermore, the distraction scoring model integrates the focus deviation magnitude, deviation change rate, the coupling term of deviation and change rate, and the regularization penalty term, and dynamically adjusts the judgment threshold through trajectory geometric features.

[0013] Furthermore, the intensity, duration, and form of the visual correction stimulus are adjusted according to the intensity index of the distraction event, which is determined by deviation, trend of change, rate of change of trajectory direction, and curvature.

[0014] Furthermore, the detection of distraction events also includes a minimum duration determination; instantaneous deviations shorter than this duration do not trigger intervention events.

[0015] Furthermore, the visual correction stimulus is selected from a pre-set visual template library, including one or more of the following: a ring-shaped bright pulse, a short flash of the target point, or a local ripple of the trajectory.

[0016] Furthermore, the difficulty coefficient is adjusted based on the difference between the distraction frequency and average deviation of the current cycle and the target value, and the difficulty scaling factor for the next cycle is calculated through linear combination.

[0017] A second aspect of the invention provides a visual tracking-based attention training device, the device comprising:

[0018] The data acquisition module is used to collect the user's real-time eye movement data, which includes the coordinates of the gaze point and the eye movement speed, and initializes the target trajectory of the training task and calculates the focus deviation between the current gaze point and the target trajectory.

[0019] The trajectory update module is used to adjust the target trajectory based on the focus deviation through nonlinear mapping to generate a dynamic trajectory;

[0020] The event detection module is used to construct a multi-feature fusion distraction scoring model based on focus deviation and its changing trend, combined with the geometric features of dynamic trajectories, to detect and label distraction events.

[0021] The correction feedback module is used to immediately execute adaptive visual correction stimuli at the location of the distraction event when a distraction event is detected, guiding the user back to the task; after the training cycle ends, the task difficulty coefficient of the next cycle is dynamically adjusted according to the frequency and average deviation of the distraction events in this cycle.

[0022] The beneficial technical effects of the present invention are at least as follows:

[0023] To address the aforementioned problems, this invention provides a visual tracking-based attention training method and apparatus. By collecting and processing real-time user gaze data, it dynamically generates a target trajectory matching the user's attention state and calculates a compact state vector containing deviation amplitude, trend of change, and trajectory geometric features. Upon detecting a distraction event matching a characteristic pattern, it immediately executes an intensity-adaptive visual stimulus at the event location to guide the user back to the task. At the end of the training cycle, the task difficulty coefficient for the next cycle is adjusted based on the frequency and mean characteristic value of distraction events in that cycle, ensuring training remains within the user's ability margin. This solution organically integrates trajectory adaptation, accurate detection, immediate correction, and personalized difficulty adjustment, forming a dual closed loop of data and control. This not only improves the real-time performance and accuracy of attention training but also enhances the adaptability and quantifiable effectiveness of long-term training. Attached Figure Description

[0024] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0025] Figure 1 This is a flowchart of a focus training method based on visual tracking according to the present invention.

[0026] Figure 2 This is a module framework diagram of a focus training device based on visual tracking according to the present invention. Detailed Implementation

[0027] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0028] In one or more embodiments, such as Figure 1 As shown, a focus training method based on visual tracking is disclosed, the method comprising the following:

[0029] S1: Collect the user's real-time eye movement data, which includes the coordinates of the gaze point and the eye movement speed, and initialize the target trajectory of the training task, and calculate the focus deviation between the current gaze point and the target trajectory;

[0030] Specifically, the goal of this step is to collect core training data through the system's perception and initialization modules and generate an initial trajectory baseline for the task. Within the same coordinate reference system, the deviation value of the current user's focus state is calculated, and the task trajectory is dynamically updated based on this deviation value to obtain the dynamic trajectory T′ used for subsequent distraction detection. t This process needs to be executed with extremely low latency during the training run to ensure a good user experience and effective training.

[0031] The initial input for this solution comes from hardware acquisition and task initialization processes, including:

[0032] gaze point coordinates P t (x, y): Collected by a calibrated eye-tracking device, such as a desktop eye tracker based on an infrared light source and a high-speed camera, or a head-mounted display device with eye-tracking capabilities. The device internally obtains the two-dimensional gaze coordinates in the screen reference frame through continuous frame pupil center detection and gaze vector calculation.

[0033] Eye speed v: Calculated by dividing the positional change of adjacent frames by the time interval using the same eye-tracking device.

[0034] Mission target trajectory T t The task generation module constructs the data during the initialization of the training cycle. For example, it can be a closed curve that moves at a constant speed on the screen or a sequence of virtual target points that move along a specific path. All trajectory point coordinates use the same reference frame as the eye-tracking data to ensure direct comparison.

[0035] First, through the current gaze point P t With the target trajectory T t Calculate the focus deviation d at the target point coordinates corresponding to the same timestamp. t :

[0036]

[0037] in, and These are the horizontal and vertical coordinates of the gaze point, respectively. and These are the coordinates of the points corresponding to the target trajectory. Both are mapped to the same screen coordinate system before the task begins using a nine-point calibration method. For example, if P... t For (850, 420) pixels, T t If the value is (870, 415) pixels, then d t Approximately 22 pixels, indicating a slight offset between the gaze position and the target.

[0038] Furthermore, d t Input a nonlinear mapping function with saturation control to convert the deviation value into the trajectory adjustment magnitude ΔT. t :

[0039]

[0040] Where, ΔT t Let k be the adjustment range of the trajectory at the current moment, k be the basic amplification factor (experimentally calibrated), and α be an adjustment parameter used to limit the adjustment amount to tend towards stability when there is a large deviation. When updating the trajectory, the adjustment range is increased along P. t Relative to T t The direction applied to T t At the current point, generate the dynamic trajectory T′ t For example, when the gaze point is above and to the left of the target point, the trajectory is slightly adjusted to the upper left, requiring the user to adjust their gaze position to re-follow the target. In a 10-second training segment, this update is performed at each sampling cycle (e.g., 120Hz refresh rate), resulting in continuous and personalized trajectory changes.

[0041] By directly translating the user's real-time focus state into changes in the task environment, dynamic difficulty adjustment is achieved; nonlinear mapping ensures sensitivity with small deviations and controllability with large deviations, enabling the system to capture minute fluctuations in focus while preventing the task from becoming unachievable.

[0042] S2: Based on the aforementioned focus deviation, the target trajectory is adjusted through nonlinear mapping to generate a dynamic trajectory;

[0043] Specifically, this step directly inherits the dynamic trajectory T′ output in the previous step. t With focus deviation d tIn our patented scenario, these two quantities are the core descriptions of "real-time task form" and "user's focus state." The core objective of this step is to ensure that, within the current training cycle, the task trajectory not only changes with the user's level of focus but also introduces a dynamic response mechanism in the trajectory adjustment that better aligns with the characteristics of focus training. For example, when the user continuously deviates from the target, the unpredictability of the target's movement should be gradually increased to stimulate their concentration; when the user quickly catches up with the target, a smooth transition is needed to avoid a sudden loss of challenge. To achieve this goal, this step introduces a focus fluctuation adjustment term and a stability constraint term into the core calculation of trajectory adjustment, ensuring that the trajectory evolution simultaneously considers both stimulation and controllability, thereby creating a continuous and effective intervention effect in focus training.

[0044] To better align the trajectory changes with the goals of focus training, this step first calculates the trend of deviation changes, i.e., the short-window average deviation change rate r. t Unlike ordinary difference calculations, we introduce a decay factor here to gradually reduce the influence of earlier deviation values ​​on the current trend, thereby highlighting recent focused fluctuations:

[0045]

[0046] Among them, w i =β i The weights are exponentially decaying (0 < β < 1) to reduce the influence of old data; n is the short window length (typically 3–5 sampling periods), Δt is the sampling interval, and sat(·) is the limiting function to prevent uncontrollable trajectory jumps caused by extreme fluctuations. This design enables the system to quickly capture recent trends in focus and reduce historical noise interference.

[0047] After calculating r t After that, we will d t With r t By fusing the data, the trajectory step size scaling factor γ is obtained. t Unlike conventional proportion control, this introduces two additional elements specifically designed for focus training:

[0048] Fluctuation adjustment term: when r t When the absolute value of the value is large, i.e. when the user's focus changes rapidly, sinusoidal modulation is added to the step size calculation to create a slight trajectory direction perturbation, simulating the dynamic interference of distracting environments in reality.

[0049] Stationarity constraint term: when d t When the step size is small, a quadratic decay function is used to suppress the increase in step size, preventing the user from losing a sense of accomplishment when the target suddenly accelerates when the user is close to the target.

[0050] Taking all these factors into account, the formula for the trajectory scaling factor is:

[0051]

[0052] Wherein: g1 is the maximum step size adjustment amplitude coefficient (experimental calibration); g2 is the sensitivity coefficient (control response change rate); λ∈[0,1] is the fusion weight, balancing the role of deviation and change rate in adjustment; η is the fluctuation adjustment amplitude coefficient, used to control the sinusoidal disturbance intensity; ω is the disturbance frequency coefficient, matched with the training task refresh rate; μ is the stationarity constraint strength coefficient; v is the stationarity constraint decay rate coefficient; the above parameters are set during training initialization and can be adjusted according to different groups (such as children, professional athletes).

[0053] In the actual trajectory update, T′ t Using the local tangential direction as a reference, multiply the nominal step size by γ. t After applying the new trajectory point T″, a new trajectory point is obtained. t To ensure a smooth trajectory, if γ t If the change in adjacent cycles exceeds a preset threshold, spline smoothing is applied to the adjustment of the current cycle to avoid sudden changes in the trajectory that are not physiologically followable.

[0054] S3: Based on focus deviation and its changing trend, and combined with the geometric features of dynamic trajectories, a multi-feature fusion distraction scoring model is constructed to detect and label distraction events.

[0055] Specifically, this step inherits all the outputs from the previous step—the quadratic dynamic trajectory T″. t With the minimum state eigenvector V t =[d t ,r t —In the specific scenario of attention training, this approach achieves high-precision distraction detection and event labeling based on dynamic trajectories. Unlike conventional visual deviation judgment, this solution emphasizes utilizing the compact features that have already integrated "attention deviation amplitude" and "change trend" in the previous step, while combining the geometric state of the trajectory itself (such as local direction change rate and curvature) for multi-dimensional judgment. This can suppress two common types of misjudgment risks in training tasks:

[0056] When the target trajectory makes a sharp turn or accelerates, the system can tolerate a certain degree of gaze deviation without being incorrectly flagged as distraction.

[0057] When the target trajectory is running smoothly, the system is more sensitive to gaze deviation and can trigger event recording at the initial stage of distraction, thereby guiding the user back to the training state in a timely manner.

[0058] This step first calculates the local geometric features of the trajectory, including the rate of change of direction θ. t With curvature κ t The rate of change of direction is measured by T″. tThe angle between the tangential vectors of the two most recent frames is obtained, and the curvature is obtained using the three-point method: with the current point T″ as the reference point. t and the sampling points T″ before and after. t-δ 、T″ t+δ We calculate the approximate radius of the arc and take its reciprocal. To adapt to the visual dynamics of the training scene, we introduce a shape sensitivity function f. geom To adjust the distraction detection threshold:

[0059]

[0060] Where, ρ θ ρ is the enhancement coefficient of the direction change rate on the threshold (reducing sensitivity when the trajectory direction changes greatly). κ θ is the sensitivity suppression coefficient when the curvature exceeds the threshold. max With κ max These are the normalized upper limits for the rate of change of direction and curvature, respectively. The special feature of this formula is that it directly quantifies the trajectory geometric features into a dynamic adjustment factor for the distraction threshold, enabling the detection to adapt to the target's motion pattern.

[0061] After obtaining f geom Then, a comprehensive distraction score S was constructed. t :

[0062]

[0063] The term consists of three parts: the first is the deviation magnitude normalization term, with weight α1 determining its impact on the scoring; the second is the deviation change rate normalization term, with weight α2 determining its impact; and the third is the deviation trend coupling term. This reflects the interaction between deviation and rate of change: when a user has a large deviation and continues to deviate, this item significantly increases the distraction score; the q in the denominator is a small constant to prevent d t Too small a value leads to instability. The fourth term is the regularization penalty term, λ. reg Controlling the penalty intensity for large, persistent deviations helps smooth the scoring curve and suppress misjudgments caused by occasional peaks.

[0064] When S t ·f geom ≥S th And the duration exceeds the minimum duration τ min At that time, the system generates a distraction event E. t And record the type (persistent / transient), start time, and duration. If S t In τ min If the price falls back within the range, it is considered a momentary deviation and does not trigger an intervention event.

[0065] S4: Upon detecting a distraction event, immediately execute adaptive visual correction stimuli at the location of the distraction event to guide the user back to the task; after the training cycle ends, dynamically adjust the task difficulty coefficient for the next cycle based on the frequency and average deviation of distraction events in this cycle.

[0066] Specifically, this step only inherits the distraction event E output from the previous step. t Snapshot of key detection volume Two things need to be accomplished: First, when an event is triggered, a visible visual correction stimulus is applied at the target reference point to quickly bring attention back to the main training line; second, at the end of the training cycle, the difficulty scaling factor for the next cycle is calculated based on the statistics of the distraction characteristics of this cycle.

[0067] Real-time correction based on the target reference point A of the event frame t Centered on the system, templates are selected from the system's pre-set visual template library (e.g., "ring-shaped bright pulse", "short-term flashing of target point", "local ripples of trajectory"), and parameters such as intensity, radius / scale, duration, and frequency are assigned according to the event intensity index I. t Adaptive settings. To balance "rapid recovery" and "avoiding excessive interruption," I t It combines smoothing bias, divergence trend, and trajectory pattern, while also attenuating ongoing events. The formula is as follows:

[0068]

[0069] Where, β d ,β r ,β θ ,β κ For weighting coefficients (parameter table configuration), d ref ,r ref ,θ ref ,κ ref φ is the normalized reference value. t ∈[0,1] represents the "duration percentage" of the event, determined by E t The duration of the stimulus is converted to a set upper limit, and it increases linearly with the duration to suppress prolonged high-intensity stimulation. The key point of this formula is: when the user continues to diverge (r t >0) or the trajectory direction changes significantly (|θ) t When the trajectory itself has a high curvature (κ), the stimulus intensity is increased; while when the trajectory itself has a high curvature (κ), the stimulus intensity is increased. t (Exceeding the threshold), automatically weakening the stimulus to avoid misinterpreting "necessary fixation due to a sharp turn in the trajectory" as strong distraction. In actual implementation, I... t By using piecewise linear mapping to the value range of the template parameter vector (such as intensity, radius, duration, and frequency), the center location is A. tThis is linked to the event type: persistent events are given longer duration and lower frequency spread pulses, while transient events use short, high-contrast flashes. For example: when I... t For events occurring in the mid-to-high intensity range and being continuous, select "ring-shaped high-brightness pulse," setting the intensity to the upper half of the range and the radius to I. t Increase, duration should be moderately high; when I t For low-level and transient conditions, select "short-term flashing of target point", and take the lower half of the interval for intensity and duration to avoid excessive intervention.

[0070] After the training period ends, the number of events and representativeness deviation within this period are counted to obtain the distraction frequency F. e With average smoothing deviation (Regarding each event) (Weighted average based on event duration). The trajectory difficulty scaling factor for the next cycle is determined by the following formula:

[0071]

[0072] In the formula, Λ base Basic difficulty level; F target The desired target distraction frequency (corresponding to the "moderate challenge" range); η F ,η D This is the sensitivity coefficient. If there is excessive distraction in this cycle (F... e >F target ) or the average deviation is too large ( If the difficulty increases, the difficulty will be appropriately reduced in the next cycle; conversely, it will be increased. This update takes effect on a cycle-by-cycle basis and will not alter the control law structure within the cycle, ensuring stable operation.

[0073] In one or more embodiments, such as Figure 2 As shown, a visual tracking-based attention training device is disclosed, the device comprising:

[0074] The data acquisition module is used to collect the user's real-time eye movement data, which includes the coordinates of the gaze point and the eye movement speed, and initializes the target trajectory of the training task and calculates the focus deviation between the current gaze point and the target trajectory.

[0075] The trajectory update module is used to adjust the target trajectory based on the focus deviation through nonlinear mapping to generate a dynamic trajectory;

[0076] The event detection module is used to construct a multi-feature fusion distraction scoring model based on focus deviation and its changing trend, combined with the geometric features of dynamic trajectories, to detect and label distraction events.

[0077] The correction feedback module is used to immediately execute adaptive visual correction stimuli at the location of the distraction event when a distraction event is detected, guiding the user back to the task; after the training cycle ends, the task difficulty coefficient of the next cycle is dynamically adjusted according to the frequency and average deviation of the distraction events in this cycle.

[0078] It is worth noting that the specific workflow of the attention training device based on visual tracking provided in this embodiment of the invention is the same as that of the attention training method based on visual tracking described in the above embodiment, and will not be repeated here.

[0079] This invention also provides a visual tracking-based attention training device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps described in the above-described visual tracking-based attention training method embodiment, for example... Figure 1 The steps S1 to S4 described above; or, when the processor executes the computer program, it implements the functions of each module in the above-described device embodiments.

[0080] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the vision-tracking-based attention training device.

[0081] The aforementioned vision-tracking-based attention training device can be a desktop computer, laptop, handheld computer, or cloud server, among other computing devices. This device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the vision-tracking-based attention training device may also include input / output devices, network access devices, and buses.

[0082] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the vision-tracking-based attention training device, connecting all parts of the device via various interfaces and lines.

[0083] The memory can be used to store the computer program and / or modules. The processor implements various functions of the vision-tracking-based attention training device by running or executing the computer program and / or modules stored in the memory, and by calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating device, at least one application program required for a function, etc.; the data storage area may store data created based on the operation of the air conditioner controller, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart memory card (SMC), secure digital card (SD), flash memory card, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0084] The module integrated into the visual tracking-based attention training device, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0085] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0086] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A focus training method based on visual tracking, characterized in that, The method includes: Collect real-time eye movement data of the user, including gaze point coordinates and eye movement velocity, initialize the target trajectory of the training task, and calculate the focus deviation between the current gaze point and the target trajectory; Based on the attention deviation, the target trajectory is adjusted through nonlinear mapping to generate a dynamic trajectory. The steps of adjusting the target trajectory include: calculating the trajectory adjustment amplitude according to the attention deviation, and limiting the upper limit of the adjustment amount through a saturation control function; applying the adjustment amplitude along the direction of the gaze point relative to the target point to generate a continuously updated dynamic trajectory; the dynamic trajectory generation process also introduces an attention fluctuation adjustment term and a stability constraint term to make the trajectory change take into account both stimulation and controllability. Based on focus deviation and its changing trend, and combined with the geometric features of dynamic trajectories, a multi-feature fusion distraction scoring model is constructed to detect and label distraction events. The distraction scoring model integrates focus deviation amplitude, deviation change rate, the coupling term of deviation and change rate, and regularization penalty term, and dynamically adjusts the judgment threshold through trajectory geometric features. Upon detecting a distraction event, an adaptive visual correction stimulus is immediately applied at the location of the distraction event to guide the user back to the task. After the training cycle ends, the task difficulty coefficient for the next cycle is dynamically adjusted based on the frequency and average deviation of the distraction events in this cycle. The intensity, duration, and presentation of the visual correction stimulus are adjusted according to the intensity index of the distraction event, which is determined by the deviation, trend of change, rate of change of trajectory direction, and curvature.

2. The attention training method based on visual tracking according to claim 1, characterized in that, The detection of distraction events also includes a minimum duration determination; instantaneous deviations shorter than this duration do not trigger an intervention event.

3. The attention training method based on visual tracking according to claim 1, characterized in that, The visual correction stimulus is selected from a preset visual template library, including one or more of the following: a ring-shaped bright pulse, a short-term flash of the target point, or a local ripple of the trajectory.

4. The attention training method based on visual tracking according to claim 1, characterized in that, The difficulty coefficient is adjusted based on the difference between the distraction frequency and average deviation of the current cycle and the target value, and the difficulty scaling factor for the next cycle is calculated by linear combination.

5. A visual tracking-based attention training device for implementing the attention training method as described in any one of claims 1 to 4, characterized in that, The device includes: The data acquisition module is used to collect the user's real-time eye movement data, which includes the coordinates of the gaze point and the eye movement speed, and initializes the target trajectory of the training task and calculates the focus deviation between the current gaze point and the target trajectory. The trajectory update module is used to adjust the target trajectory based on the focus deviation through nonlinear mapping to generate a dynamic trajectory; The event detection module is used to construct a multi-feature fusion distraction scoring model based on focus deviation and its changing trend, combined with the geometric features of dynamic trajectories, to detect and label distraction events. The correction feedback module is used to immediately execute adaptive visual correction stimuli at the location of the distraction event when a distraction event is detected, guiding the user back to the task; after the training cycle ends, the task difficulty coefficient of the next cycle is dynamically adjusted according to the frequency and average deviation of the distraction events in this cycle.

Images