Children concentration training method and training system

Abstract

The invention provides a child concentration training method and training system. The training method comprises the following steps: loading a virtual scene for child concentration training, wherein the virtual scene comprises a task scene and a virtual pet; controlling the virtual pet to display pet behaviors when the child executes the training task, and continuously acquiring multiple groups of multi-modal physiological and behavior data when the child executes the training task according to a preset time interval; based on the multi-modal physiological and behavior data of the current group, acquiring the emotion activation degree, the fixation stability and the red region drift index of the child, and inputting the emotion activation degree, the fixation stability and the red region drift index into the trained decision tree model to obtain a current cognitive load evaluation result; performing dynamic task adjustment and virtual pet behavior adjustment based on the current cognitive load evaluation result; and obtaining a concentration training result. According to the training method provided by the invention, teaching is carried out according to materials, the interestingness of children training can be increased, the enthusiasm of children training can be aroused, and children can complete tasks more happily.

Description

Technical Field

[0001] The present invention belongs to the technical field of virtual reality interaction technology, and specifically relates to a children's concentration training method and training system. Background Art

[0002] At present, the training methods for children's concentration mainly include static content such as picture card recognition, task punching, and video guidance. They lack interactivity and fun, and are easy to cause boredom, especially for young children, resulting in poor training continuity. In addition, many training systems use a unified training path and content, ignoring individual differences in cognitive ability, emotional state, attention span, etc., making it difficult to teach students in accordance with their aptitude. Individual training feedback often relies only on task completion (such as whether it is completed, completion time, etc.), and cannot reflect the child's cognitive load level, emotional changes, and attention fluctuations during the training process. Traditional assessment methods are mostly offline paper-and-pencil tests or expert observation and scoring. Not only are the assessment cycles long and real-time poor, but there are also problems such as strong subjectivity and low data granularity. Especially in the context of the increasing popularity of distance learning and home training, there is a lack of real-time, objective, and high-frequency data collection methods, making it difficult to form an effective training closed loop. Summary of the Invention

[0003] The purpose of the present invention is to fully consider the differences among children in cognitive ability, emotional state, and attention span, and to provide a method for training children's concentration that is tailored to their aptitude. During the training process, the method controls a virtual pet to accompany the child in training, and the behavior of the virtual pet is controlled based on the cognitive load state and concentration of the child during the training process. This can increase the fun of children's training and mobilize children's enthusiasm for training, so that children can complete tasks more happily.

[0004] In a first aspect, the present invention provides a method for training children's concentration, which is applied to a wearable device loaded with a virtual pet interactive platform, and the training method includes the following steps: step S10, receiving a loading instruction, and loading a virtual scene for children's concentration training based on the loading instruction, wherein the virtual scene includes a task scene corresponding to the training task and a virtual pet for accompanying the child in training; step S20, controlling the virtual pet to display pet behavior when the child performs the training task, and continuously obtaining multiple groups of multimodal physiological and behavioral data of the child in performing the training task at preset time intervals, each group of multimodal physiological and behavioral data including eye movement trajectory data and facial images; step S30, based on the multimodal physiological and behavioral data of the current group, Data is collected to obtain the child's emotional activation level, gaze stability and red zone drift index, and the child's emotional activation level, gaze stability and red zone drift index are input into the trained decision tree model to obtain the current cognitive load assessment result, which includes the cognitive load status and concentration level; step S40, the current cognitive load assessment result is used as the updated cognitive load assessment result, and dynamic task regulation and dynamic adjustment of the pet behavior displayed by the virtual pet when the child performs the next training task are performed according to the updated cognitive load assessment result; step S50, when the training time reaches the target training time, the virtual scene is closed, the concentration training is evaluated, and the concentration training result is obtained.

[0005] In a specific embodiment, the training method also includes a step before step S10: obtaining the child's basic file information, the basic file information including the attention level baseline, physiological parameter range and user basic information; the task scenario loaded in step S10 is determined based on the basic file information.

[0006] In a specific embodiment, the step of obtaining the degree of emotional activation includes: step (1), preprocessing the facial images collected within the preset time interval to obtain preprocessed facial images, wherein the preprocessing includes cropping the facial area and image normalization, wherein the number of facial images is one frame or multiple frames; step (2), extracting the features of each frame of the preprocessed facial image to obtain at least one facial feature set corresponding to each frame of the preprocessed facial image, and inputting each facial feature set into an expression classification model to obtain a probability distribution result of the expression classification corresponding to each frame of the facial image, wherein the probability distribution result of the expression classification includes the expression classification name and the probability of the expression classification, and the expression classification name includes happiness, surprise, anger, disgust, sadness, fear and neutrality; step (3), based on the preset typical emotional activation degree corresponding to each type of expression and the probability distribution result of the expression classification corresponding to each frame of the facial image, obtaining the emotional activation degree corresponding to each frame of the facial image; step (4), obtaining the average value of the emotional activation degrees corresponding to all facial images, and taking the average value of the emotional activation degrees as the emotional activation degree of the child.

[0007] In a specific embodiment, the eye movement trajectory data includes the coordinates of the gaze point, the gaze time, and the scanning path and distance, and the step of obtaining the gaze stability includes: step (1), projecting each gaze point coordinate into the interface coordinate system of the task scene, and forming a gaze point cloud map based on the gaze time and the scanning path and distance; step (2), dividing the gaze point cloud map into two-dimensional grids, and counting the cumulative gaze time for each grid, using Gaussian kernel density estimation to fuzzy weight the point cloud, and generating a red zone migration map with a gradient color of the gaze point; step (3), based on the key cognitive sites preset in the task scene and the red zone migration map, calculating the gaze stability according to a first formula, wherein the first formula is:

[0008]

[0009] Among them, FS is gaze stability, T target is the total duration of fixation on the key cognitive point, T total is the total duration of all fixations.

[0010] In a specific embodiment, the red zone drift index is calculated based on the eye movement trajectory data according to a second formula, wherein the second formula is:

[0011]

[0012] DI is the red zone drift index, (x i ,y i) is the coordinate of the i-th fixation point, and N is the total number of fixations.

[0013] In a specific embodiment, the concentration training results include a cognitive load curve and a red zone migration diagram generated in real time during the training process, and a visual concentration trend report output after the training is completed.

[0014] In a specific embodiment, the training method also includes storing the training records and concentration training results of each training session, and dynamically adjusting the task type of the next training task and the visual layout and interaction frequency of the task scene based on the historical training records and the red zone migration map obtained from the current training, so as to load suitable task scenes for children with different cognitive styles.

[0015] In a specific embodiment, the pet behavior includes prompting interaction behavior for guiding or prompting the current task goal, rewarding interaction behavior for providing visual and emotional encouragement after completing the task, and accompanying interaction behavior for providing emotional comfort and non-task oriented interaction.

[0016] In a specific embodiment, the concentration level includes four levels: high, medium, low and very low. Step S40 also includes comparing the current concentration level with the previous concentration level, and when the concentration level drops directly from high to low, adjusting the virtual pet to display the pet behavior corresponding to the short-term decrease.

[0017] In a second aspect, the present invention further provides a children's concentration training system, which is applied to a wearable device loaded with a virtual pet interaction platform, and the training system includes: a loading module for receiving a loading instruction, and loading a virtual scene for children's concentration training based on the loading instruction, wherein the virtual scene includes a task scene corresponding to the training task and a virtual pet for accompanying the child in training; a first acquisition module for controlling the virtual pet to display pet behavior when the child performs the training task, and continuously acquiring multiple groups of multimodal physiological and behavioral data of the child in performing the training task at preset time intervals, each group of multimodal physiological and behavioral data including eye movement trajectory data and facial images; a second acquisition module for acquiring data based on the multimodal physiological and behavioral data of the current group The method comprises the following steps: using the data from the training program to obtain the child's emotional activation level, gaze stability and red zone drift index, and inputting the child's emotional activation level, gaze stability and red zone drift index into the trained decision tree model to obtain the current cognitive load assessment result, which includes the cognitive load status and concentration level; an adjustment module, which is used to use the current cognitive load assessment result as the updated cognitive load assessment result, and dynamically control the task according to the updated cognitive load assessment result, and dynamically adjust the pet behavior displayed by the virtual pet when the child performs the next training task; a third acquisition module, which closes the virtual scene when the training time reaches the target training time, evaluates the concentration training, and obtains the concentration training result.

[0018] The beneficial effects of the present invention include at least:

[0019] The method for training children's concentration provided by the present invention is applied to a wearable device loaded with a virtual pet interactive platform, and includes the following steps: step S10, receiving a loading instruction, and loading a virtual scene for children's concentration training based on the loading instruction, wherein the virtual scene includes a task scene corresponding to the training task and a virtual pet for accompanying the child in training; step S20, controlling the virtual pet to display pet behavior when the child performs the training task, and continuously obtaining multiple groups of multimodal physiological and behavioral data of the child in performing the training task at preset time intervals, each group of the multimodal physiological and behavioral data including eye movement trajectory data and facial images; step S30, based on the multimodal physiological and behavioral data of the current group, obtaining the child's emotional activation level, gaze stability and red zone drift index, and inputting the child's emotional activation level, gaze stability and red zone drift index into the trained decision tree model to obtain the current cognitive load assessment result. The cognitive load assessment result includes the cognitive load status and the concentration level; step S40, taking the current cognitive load assessment result as the updated cognitive load assessment result, and dynamically controlling the task and dynamically adjusting the pet behavior displayed by the virtual pet when the child performs the next training task according to the updated cognitive load assessment result; step S50, closing the virtual scene when the training time reaches the target training time, evaluating the concentration training, and obtaining the concentration training result; the concentration training method provided by the present invention creatively loads a virtual pet to accompany the child for training, and the behavior of the virtual pet is timely adjusted based on the cognitive load status and concentration of the child during the training process. In this way, on the one hand, the fun of the children's training process can be increased, which helps to assist the children in completing the training. On the other hand, the concentration training method fully considers the differences among children in cognitive ability, emotional state, and attention span, and can improve the effect of concentration training.

[0020] In addition to the above-described objects, features and advantages, the present invention has other objects, features and advantages. The present invention will be further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of a step flow chart of a method for training children's concentration provided by one embodiment of the present invention;

[0022] Figure 2 This is a visual design diagram of a virtual pet character in one embodiment of the present invention;

[0023] Figure 3 This is a red zone migration map constructed based on the task type of the training task, Find the Difference game, as an example;

[0024] Figure 4A cognitive load curve in a concentration training result displayed as a visualization graph provided by an embodiment of the present invention;

[0025] Figure 5 A training trend report of the concentration training results presented in a visual graph provided by an embodiment of the present invention;

[0026] Figure 6 This is a module diagram of a children's concentration training system provided by another embodiment of the present invention. DETAILED DESCRIPTION

[0027] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0028] Concentration, also known as attention, refers to the psychological state of a person when they are focused on a particular object or activity. Under normal circumstances, concentration directs one's mental activity toward a specific object, selectively accepting certain information while suppressing other activities and information, and focusing all mental energy on the desired object. Therefore, good concentration helps improve work and study efficiency.

[0029] According to a first aspect of the present invention, the present invention provides a method for training children's concentration, which is applied to a wearable device loaded with a virtual pet interactive platform. The present invention increases the fun of children's training by loading a virtual pet to accompany the child in training according to predetermined pet behaviors while the child is performing a training task. On the other hand, the pet behavior of the virtual pet can also be used to mobilize the enthusiasm of the child for training, so that the child can complete the task more happily.

[0030] In the present invention, the virtual pet interaction platform may include the following: a virtual pet core engine: including an animation rendering module, a behavior decision module, and an emotion expression module; an interface adaptation system: automatically adjusting the visual layout according to different screen sizes and interaction methods; a voice / touch input interface: supporting multimodal human-computer interaction, including voice command recognition, gesture / touch feedback, etc.; a synchronous task system: enabling real-time linkage between virtual pet behavior and cognitive training task processes.

[0031] In the present invention, the wearable device can be a wearable device specifically for children, or it can be smart glasses.

[0032] In an optional embodiment of the present invention, the wearable device includes a camera, a microphone, a sensor module and a data processing module. The camera is used to collect facial images and body movements of the child, the microphone is used for calls, the sensor module is used to collect physiological parameters, and may include a heart rate belt, an EEG headband and a skin inductor, etc. The data processing module is used to store and process the collected data.

[0033] In an optional embodiment of the present invention, the wearable device can also input data through touch input or voice input.

[0034] In another optional embodiment of the present invention, the sensor module may further include a brain wave biosensor.

[0035] See also Figure 1 , the training method comprises the following steps:

[0036] Step S10: receiving a loading instruction, and loading a virtual scene for children's concentration training based on the loading instruction, wherein the virtual scene includes a task scene corresponding to the training task and a virtual pet for accompanying the children in training.

[0037] In an optional embodiment, the types of training tasks include puzzles, image sorting, path tracing, matching tasks, spot the differences, and guided mini-games.

[0038] In the present invention, a virtual pet is added to the task scenario. The virtual pet is not only a companion element, but also an intelligent interaction engine. It recognizes the child’s current cognitive state and dynamically adjusts behavior and conversation style to achieve "education through entertainment, guided training, and real-time intervention."

[0039] See also Figure 2 , Figure 2 This is a visual design diagram of a virtual pet character in an embodiment of the present invention.

[0040] Step S20: Control the virtual pet to display pet behavior when the child performs the training task, and continuously obtain multiple sets of multimodal physiological and behavioral data of the child when performing the training task at preset time intervals, each set of the multimodal physiological and behavioral data including eye movement trajectory data and facial images.

[0041] In an optional embodiment, the pet behavior includes prompting interaction behavior for guiding or prompting the current task goal, rewarding interaction behavior for providing visual and emotional encouragement after completing the task, and companionship interaction behavior for providing emotional comfort and non-task oriented interaction.

[0042] In the present invention, the interaction mechanism between virtual pets and children is the key bridge connecting cognitive assessment and personalized intervention in the technical solution of the present invention. It transforms programmed cognitive feedback into a warm, vivid, and game-like interactive experience with virtual pets, thereby enhancing children's participation and training stickiness.

[0043] In an optional embodiment, the eye movement trajectory data includes gaze point coordinates, gaze duration, and scan path and frequency.

[0044] In the present invention, the processing algorithm of the eye movement trajectory data includes: facial key point recognition + iris tracking to achieve lightweight eye movement tracking; using a filtering algorithm (such as Kalman Filter) to remove micro-jitter and noise, which are all existing technologies and the present invention does not make any improvements on them.

[0045] In an optional embodiment, the multimodal physiological and behavioral data also includes operational behavior data, and the operational behavior data includes task behaviors of children performing training tasks, interactive behaviors of children and pets, and interactive behavior characteristic data corresponding to the interactive behaviors.

[0046] In the present invention, the operational behavior data and the eye movement trajectory data can be combined to analyze the "eye-hand consistency" for high-level cognitive control level estimation. It can also be determined based on the operational behavior data whether there is task avoidance or anxious procrastination behavior. The high-level cognitive control level estimation results and whether there is task avoidance or anxious procrastination behavior during the training process can be output as part of the concentration training results to facilitate parents to have a more comprehensive understanding of their children's concentration training status.

[0047] In the present invention, the task behavior refers to the actions of actively controlling, modifying, activating, and providing feedback on interface elements or task objects when children perform training tasks, including clicking, dragging, zooming in / out, sliding, long pressing, double-clicking, dragging, input, interactive triggering, etc.

[0048] In the present invention, the interactive task refers to the process of children performing training tasks and controlling and giving feedback to virtual pets, including clicking, dragging, sliding, inputting, etc. For example, clicking a reward button, dragging candy to feed a virtual pet, sliding the interface to switch pet categories, inputting a pet nickname, etc.

[0049] The interactive behavior characteristic data refers to data indicators used to describe the details of interactive behaviors, including action categories, frequency, time, success rate, etc.

[0050] For ease of understanding, for example, if the interactive behavior is dragging, the data indicators corresponding to the dragging behavior include the starting point coordinates, the vertical point coordinates, the dragging time, the trajectory path, and whether the target is successfully triggered.

[0051] Step S30: Based on the multimodal physiological and behavioral data of the current group, the child's emotional activation level, gaze stability, and red zone drift index are obtained, and the child's emotional activation level, gaze stability, and red zone drift index are input into the trained decision tree model to obtain the current cognitive load assessment result, which includes the cognitive load status and concentration level.

[0052] In the present invention, the multimodal physiological and behavioral data of the current group can be understood as the updated multimodal physiological and behavioral data or the multimodal physiological and behavioral data acquired at the current time. For ease of understanding, for example, assuming the total training time is 5 minutes and the current training time is 2 minutes, the multimodal physiological and behavioral data of the current group is the multimodal physiological and behavioral data corresponding to the training time of 2 minutes.

[0053] In the present invention, the cognitive load state and the concentration level are negatively correlated, the cognitive load state includes low load, medium load, high load and extremely high load, and the concentration level includes four levels: high, medium, low and extremely low.

[0054] The step of obtaining the emotional activation degree includes:

[0055] Step (1) preprocessing the facial images collected within the preset time interval to obtain preprocessed facial images, wherein the preprocessing includes cropping the facial area and image standardization, wherein the number of the facial images is one frame or multiple frames.

[0056] In the present invention, the preset time interval can be set based on an empirical value. Within the preset time interval, only one frame of facial image can be collected, or two frames of facial images, three frames of facial images, or any other number of facial images can be collected. It can be understood that the shorter the preset time interval, the greater the data processing volume, and at the same time, the update of the emotional activation level is more timely.

[0057] In the present invention, the facial image acquisition frequency is at least 15 fps.

[0058] In the present invention, MTCNN or BlazeFace can be used to crop the facial area and the method provided by the prior art is used for processing, and the present invention does not make any improvements.

[0059] In the present invention, the image normalization process includes grayscale conversion and size normalization.

[0060] Step (2) extracts the features of each frame of preprocessed facial image respectively to obtain at least one facial feature set corresponding to each frame of preprocessed facial image, and inputs each facial feature set into an expression classification model to obtain a probability distribution result of the expression classification corresponding to each frame of facial image, wherein the probability distribution result of the expression classification includes an expression classification name and a probability of the expression classification, and the expression classification name includes happiness, surprise, anger, disgust, sadness, fear and neutrality.

[0061] In the present invention, each facial feature set includes 68 facial features, and facial feature extraction can be: detecting 68 key points of the face (using libraries such as Dlib, MediaPipe, etc.); or using CNN to extract deep features (such as FaceNet, ResNet50 variants).

[0062] Step (3) obtains the emotion activation degree corresponding to each facial image frame based on the preset typical emotion activation degree corresponding to each type of expression and the probability distribution result of the expression classification corresponding to each facial image frame.

[0063] In the present invention, there is a mapping relationship between each type of expression name and the typical emotion activation degree corresponding to each type of expression name, and the typical emotion activation degree corresponding to the expression can be obtained based on the expression classification name.

[0064] For ease of understanding, the mapping relationship between expression category names and corresponding typical emotion activation levels can be as follows: happy—high emotion activation level (0.7-0.9); sad—low emotion activation level (0.2-0.4); surprised—high emotion activation level (0.8-1.0); neutral—medium emotion activation level (0.4-0.5).

[0065] Step (4): obtaining an average value of the emotion activation levels corresponding to all facial images, and using the average value of the emotion activation levels as the emotion activation level of the child.

[0066] In the present invention, the emotional activation level is between 0 and 1, including 0 and 1.

[0067] For ease of understanding, let's take an example: suppose the probability distribution of expression classification for a frame of image is: happy: 70%, surprised: 20%, neutral: 10%; the corresponding typical emotion activation level settings are: happy: 0.8, surprised: 0.9, neutral: 0.5; then the emotion activation level E = 0.7×0.8+0.2×0.9+0.1×0.5=0.56+0.18+0.05=0.79.

[0068] If only this frame of image is acquired during the preset time interval, the child's emotional activation level E is 0.79. If three frames of image are acquired during the preset time interval, and the emotional activation levels of the other two frames are 0.60 and 0.84 respectively, the child's emotional activation level = (0.79+0.60+0.84) / 3 = 0.74.

[0069] The step of acquiring gaze stability comprises:

[0070] Step (1) projects the coordinates of each gaze point into the interface coordinate system of the task scene, and forms a gaze point cloud map based on the gaze time, scanning path and distance.

[0071] It is understandable that a location will be marked as a gaze point only after the time it is gazed at exceeds a preset time.

[0072] Step (2) divides the gaze point cloud map into two-dimensional grids, and counts the cumulative gaze time for each grid, uses Gaussian kernel density estimation to perform fuzzy weighting on the point cloud, and generates a red zone migration map with a gradient color of the gaze point.

[0073] Please refer to Figure 3 , Figure 3 The red zone migration map is constructed based on the task type of the find-the-difference game as an example. The center position of the fixation point on the red zone migration map corresponds to the longest fixation time, and the fixation time gradually decreases from the center to the outside.

[0074] In the present invention, the size of the small area divided by the two-dimensional grid is 32 pixels×32 pixels.

[0075] Step (3): Calculate gaze stability according to the first formula based on the key cognitive sites preset in the task scene and the red zone migration map.

[0076] In the present invention, the preset key cognitive sites can be understood as the points that children need to pay attention to when performing training tasks, and can also be understood as the correct positions for examining concentration.

[0077] The gaze stability is the ratio of the total duration of gaze on the key cognitive point to the total duration of all gaze points. The first formula is:

[0078]

[0079] Among them, FS is gaze stability, T target is the total duration of fixation on the key cognitive point, T total is the total duration of all fixations.

[0080] The higher the gaze stability value, the better the user's concentration. Otherwise, the user may be distracted or glance too much. Figure 3 ,In the find the different red zone migration map provided by an embodiment of the ,presentation, if the user focuses mainly on the difference point area in the ,find the difference task, the gaze stability is high.

[0081] For ease of understanding, let's take an example. If the number of fixation points on the task red map is 5, the number of key cognitive sites is also 5, and the 5 key cognitive sites and the 5 fixation points completely overlap, then the fixation stability is 1. If any of the 5 fixation points is not a key cognitive site, then the fixation stability is 0. If any of the 5 fixation points has 3 key cognitive sites, then the fixation stability is the ratio of the duration of the 3 pairs of fixation sites to the duration of the 5 fixation sites.

[0082] In the present invention, based on the eye movement data, the red zone drift index is calculated according to a second formula, wherein the second formula is:

[0083]

[0084] DI is the red zone drift index, (x i ,y i ) is the coordinate of the i-th fixation point, and N is the total number of fixations.

[0085] The red zone drift index is used to measure the intensity of the tendency of the gaze point to "drift" from one key area to another. A high drift index may reflect cognitive fatigue, unclear task objectives, and many environmental interferences; a low drift index represents: focused attention and stable task execution.

[0086] In the present invention, the training method of the decision tree model is specifically as follows: the X value in the model training input data is the emotional activation level, gaze stability, and red zone drift index obtained based on multimodal physiological and behavioral data collected during the children's training process; the Y value is the subjective evaluation result (cognitive load status and concentration level) of experts based on the process data of children performing tasks of different difficulty levels, corresponding behavioral videos, and facial expression data; K-fold cross-validation is used during the training process to improve the generalization ability of the model.

[0087] In this invention, by introducing the red zone migration map and decision tree model, the quantitative analysis ability of the training process can be improved.

[0088] Step S40: Using the current cognitive load assessment result as an updated cognitive load assessment result, and dynamically controlling the task and dynamically adjusting the pet behavior displayed by the virtual pet when the child performs the next training task according to the updated cognitive load assessment result.

[0089] In the present invention, dynamic task control based on the updated cognitive load assessment results can be understood as timely adjusting the task difficulty of the child's next training task according to the child's current training status. For example, when the cognitive load status is low and the concentration level is high, the task difficulty of the child's next training task can be increased or changed to a new one (the task type can be changed); when the cognitive load status is extremely high and the concentration level is extremely low, the virtual pet can be controlled to give prompts to reduce the task difficulty, etc.

[0090] In the present invention, the pet behavior in step S20 is determined based on an updated cognitive load assessment result corresponding to the child performing the training task or is an initial pet behavior set for the virtual pet by the virtual pet interaction platform.

[0091] Specifically, when the child just starts to perform the training task, the virtual pet is controlled to display the initial pet behavior; after obtaining the first cognitive load assessment result, when the child performs the next training task, the virtual pet is controlled to display the pet behavior corresponding to the first cognitive load assessment result; after obtaining the second cognitive load assessment result, that is, after the cognitive load assessment result is updated for the first time, when the child performs the next training task, the virtual pet is controlled to display the pet behavior corresponding to the second cognitive load assessment result; after obtaining the third cognitive load assessment result, that is, after the cognitive load assessment result is updated for the second time, when the child performs the next training task, the virtual pet is controlled to display the pet behavior corresponding to the third cognitive load assessment result... I will not list them one by one here.

[0092] For ease of understanding, for example, assuming that the preset time interval is 5 seconds, that is, the cognitive load assessment result is updated every 5 seconds, in the first 5 seconds of training, the pet behavior is controlled to display the initial pet behavior, in the 6th to 10th seconds of training, the pet behavior corresponds to the cognitive load assessment result corresponding to the first 5 seconds of training, in the 11th to 15th seconds of training, the pet behavior corresponds to the cognitive load assessment result corresponding to the 6th to 10th seconds of training, and in the 16th to 20th seconds of training, the pet behavior corresponds to the cognitive load assessment result corresponding to the 11th to 15th seconds of training, that is, the currently trained pet behavior is updated based on the last updated cognitive load assessment result, and the cognitive load assessment result is determined by the child's training status. In this way, the pet behavior is updated in real time based on the child's training status, so that the virtual pet can display more suitable pet behavior to accompany the child to better complete the training.

[0093] In the present invention, the pet's behavior is adjusted in a timely manner through updated cognitive load assessment results to achieve dynamic task adaptation, which can better accompany and guide children to complete training tasks and improve the concentration training effect.

[0094] In the present invention, the corresponding relationship between the cognitive load assessment results and the communication behavior and dynamic task regulation of the virtual pet is shown in Table 1.

[0095] Table 1

[0096]

[0097] It should be noted that Table 1 is merely an example of the corresponding relationship between the cognitive load assessment results and the communication behavior of the virtual pet. In other implementations, the pet behavior of the virtual pet may also be redesigned.

[0098] For easier understanding, let's take an example. Scenario 1: When a child's cognitive load is high, the pet voices, "We were a little fast. Can you take a break? You're doing great!" The pet's actions include squatting, blinking, and wagging its tail to offer comfort. Dynamic task control: The task difficulty automatically decreases, rewards are suspended, and the task switches to companionship mode.

[0099] Scenario 2: When the child's cognitive load is low, the pet voices, "Even this little thing isn't a problem for you! Come challenge my favorite game!" The pet's actions include jumping, spinning, and laughing. Dynamic task control: New tasks that increase the challenge are automatically introduced.

[0100] Scene 3: The task is completed, and the pet celebrates: popping stars, giving out candies, and upgrading skills; playing applause sound effects; providing "pet evolution" rewards - such as hats and props, to strengthen training motivation.

[0101] In an optional embodiment, step S40 further includes comparing the current concentration level with the previous concentration level, and when the concentration level drops directly from high to low, controlling the virtual pet to display pet behavior corresponding to the short-term drop.

[0102] In the present invention, the short-term reduction of the corresponding pet behavior may be: the pet briefly pauses, asks questions or makes jokes to attract attention; switches to interesting tasks, etc.

[0103] Step S50: When the training duration reaches the target training duration, the virtual scene is closed, the concentration training is evaluated, and the concentration training result is obtained.

[0104] In the present invention, the target training duration can be freely set by parents and can be set to 6 minutes, 8 minutes, 10 minutes, etc. The duration can be set based on the child's age, personality, acceptable training duration, etc.

[0105] See also Figure 4 and Figure 5 , Figure 4 and Figure 5 The results of concentration training are shown in a visual graph, among which: Figure 4 is the cognitive load curve, Figure 5 For training trend reports.

[0106] In an optional embodiment, the concentration training results include a cognitive load curve and a red zone migration diagram generated in real time during the training process, and a visual concentration trend report output after the training is completed.

[0107] In an optional embodiment, the concentration training result may also include a high-order cognitive control level assessment result, and the high-order cognitive control level assessment result is obtained based on the eye movement trajectory data and the operational behavior data.

[0108] In the present invention, the cognitive load curve is used to show how the cognitive load intensity changes over time during training. The horizontal axis is time, the vertical axis is cognitive load level, and the technical key is curve smoothing.

[0109] In the present invention, the red zone migration map is used to reflect the changing trajectory of the children's attention focus area in the task scenario.

[0110] The concentration trend report is used for review and analysis. The content of the concentration trend report includes total training time, effective average time, average cognitive load, number of attention fluctuations, red zone stability index and recommended optimization strategies.

[0111] The Red Zone Stability Index reflects the duration and stability of a user's high-focus / high-activation state during focus training or emotional management. A higher Red Zone Stability Index indicates that the user can maintain a high-focus state for a longer period of time, while maintaining a relatively stable emotional or cognitive load.

[0112] The calculation method of the red zone stability index is as follows:

[0113] 1. Define red zone conditions

[0114] Set a threshold for emotional activation, such as a red zone for emotional activation levels ≥ 0.7. Set a fixed sampling frequency, such as recording cognitive load or emotional activation levels once per second.

[0115] 2. Continuous red zone time period detection

[0116] Check the time periods that are continuously in the red zone (for example, 10 consecutive seconds in the red zone is considered a stable interval). Record the length of each stable interval, such as 8 seconds, 12 seconds, 15 seconds, etc.

[0117] 3. Calculate the degree of volatility

[0118] Within the red zone, calculate the fluctuation of the load / emotional activation value (standard deviation, maximum and minimum difference, etc.).

[0119] If the fluctuation is too large (such as exceeding the set threshold, such as the emotional activation value fluctuates by more than 0.2), points will be deducted.

[0120] 4. Comprehensive scoring

[0121] Red zone coverage = (red zone time / total training time) × 100%.

[0122] Average stable zone length = (total stable time period / number of stable segments).

[0123] The fluctuation range in the red zone = standard deviation or maximum and minimum difference.

[0124] The final red zone stability index can be calculated as follows:

[0125] Red zone stability index = (red zone coverage × 0.6) + (normalized score of average stable zone length × 0.3) + (small fluctuation score × 0.1)

[0126] Among them: the red zone coverage rate has a high weight (0.6), because stability first requires entering the red zone; the stability length has a medium weight (0.3), which encourages long-term maintenance; the small fluctuation has a small weight (0.1) to avoid excessive punishment of natural small fluctuations.

[0127] The final results were normalized to a score of 0 to 100.

[0128] For ease of understanding, let's take an example. For a 10-minute training session, the data analysis is as follows: total red zone time: 7 minutes (70% coverage); average red zone segment length: 28 seconds (the full score setting is 30 seconds, score 28 / 30≈93 points); the standard deviation of fluctuations within the red zone is less than 0.1 (very stable, full marks).

[0129] Then the red zone stability index = (70×0.6)+(93×0.3)+(100×0.1)=42+27.9+10=79.9.

[0130] That is, the red zone stability index is rounded to 80 points.

[0131] In an optional embodiment, the training method also includes storing the training records and concentration training results of each training session, and dynamically adjusting the task type of the next training task and the visual layout and interaction frequency of the task scene based on the historical training records and the red zone migration map obtained from the current training, so as to load suitable task scenes for children with different cognitive styles.

[0132] In the present invention, task type adaptation is specifically as follows: based on the task types that perform well in the training records, relevant types (such as puzzles, image sorting, path tracing, matching tasks, find the differences, etc.) are pushed first; if the user's cognitive load for a certain type of task is continuously high, it can be replaced in stages with relief tasks, such as guided mini-games.

[0133] In the present invention, the visual layout adjustment is specifically as follows: rearranging key cognitive points so that children can focus more on the training target area; visually strengthening the "low attention area", such as deepening the color and dynamically highlighting the edges; controlling the complexity of the picture to avoid cognitive overload.

[0134] In the present invention, the interaction frequency adjustment is specifically as follows: when attention fluctuates frequently, the frequency of virtual pet feedback is increased (such as voice encouragement, animation prompts); during the peak period of concentration, interference prompts are appropriately reduced to enhance immersion; a trigger mechanism is set for the "interaction cold zone" to guide children's attention.

[0135] In an optional embodiment, the training method further includes a step before step S10: obtaining basic file information of the child, the basic file information including an attention level baseline, a physiological parameter range, and basic user information; the task scenario loaded in step S10 is determined based on the basic file information.

[0136] In an optional embodiment, the user information includes age, gender, and grade or school level.

[0137] In the present invention, the basic profile information can be directly input through the wearable device, or can be input through a tablet, computer or other electronic device that is communicatively connected to the wearable device.

[0138] In the present invention, the design of training tasks for training children's concentration is related to the children's age, gender, grade, or the children's age, gender, and school level. It can be understood that multiple different training tasks are designed for children of different ages, genders, and grades / school levels, making the training more targeted.

[0139] For ease of understanding, for example, matching first-level training tasks can be designed for girls aged 2 to 4 who are not in primary school; matching second-level training tasks can be designed for boys aged 2 to 4 who are not in primary school; matching third-level training tasks can be designed for girls aged 4 to 6 who are not in primary school; matching fourth-level training tasks can be designed for boys aged 4 to 6 who are not in primary school; matching fifth-level training task sets can be designed for children aged 6 to 9 who are in grades less than 3... and we will not list them all here.

[0140] When the user is a boy aged 2 to 4 who is not yet in primary school, the task scenario loaded in step S10 is selected from the second level training task.

[0141] The method for obtaining the attention level baseline is: obtaining the average gaze time, attention duration, completion speed and accuracy of children performing baseline tasks, where the baseline task is a short-term gaze task or a basic recognition game; and obtaining the attention level baseline based on the average gaze time, attention duration, completion speed and accuracy.

[0142] In the present invention, the baseline task may be to find the differences between two pictures.

[0143] In the present invention, the attention level baseline is a value between 0 and 1, and the calculation formula is: attention level baseline = A×average gaze time + B×attention duration + C×(1-completion speed) + D×accuracy, where A+B+C+D=1.

[0144] In the present invention, A, B, C and D can be understood as weight scores.

[0145] For ease of understanding, let's take an example. Assuming A = 0.25, B = 0.25, C = 0.2, and D = 0.3, then the attention level baseline = 0.25 × average gaze time + 0.25 × attention duration + 0.2 × (1-completion speed) + 0.3 × accuracy.

[0146] It should be noted that when calculating the attention level baseline, the average fixation time, attention duration, completion speed and accuracy are all normalized and converted into data in the range of 0 to 1.

[0147] In the present invention, the physiological parameter range is collected by a sensor module.

[0148] Please refer to Figure 6 According to a second aspect of the present invention, the present invention further provides a children's concentration training system 100, which is applied to a wearable device loaded with a virtual pet interaction platform, and the training system includes:

[0149] The loading module 101 is used to receive a loading instruction and load a virtual scene for children's concentration training based on the loading instruction. The virtual scene includes a task scene corresponding to the training task and a virtual pet for accompanying the child in training.

[0150] The first acquisition module 102 is used to control the virtual pet to display pet behavior when the child performs the training task, and continuously obtain multiple sets of multimodal physiological and behavioral data of the child during the training task at preset time intervals, each set of the multimodal physiological and behavioral data including eye movement trajectory data and facial images.

[0151] The second acquisition module 103 is used to obtain the child's emotional activation level, gaze stability and red zone drift index based on the multimodal physiological and behavioral data of the current group, and input the child's emotional activation level, gaze stability and red zone drift index into the trained decision tree model to obtain the current cognitive load assessment result, which includes the cognitive load status and concentration level.

[0152] The adjustment module 104 is configured to use the current cognitive load assessment result as an updated cognitive load assessment result, and to dynamically control the task and dynamically adjust the pet behavior displayed by the virtual pet when the child performs the next training task based on the updated cognitive load assessment result.

[0153] The third acquisition module 105 closes the virtual scene when the training duration reaches the target training duration, evaluates the concentration training, and obtains the concentration training result.

[0154] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the working process of the training system 100 described above can refer to the contents of the aforementioned method embodiment and will not be repeated here.

[0155] The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, several simple deductions and substitutions can be made without departing from the concept of the present invention, and all of these should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for training children's concentration, applied to a wearable device loaded with a virtual pet interactive platform, characterized in that: The training method comprises the following steps: Step S10: receiving a loading instruction, and loading a virtual scene for children's concentration training based on the loading instruction, wherein the virtual scene includes a task scene corresponding to the training task and a virtual pet for accompanying the child in training; Step S20: controlling the virtual pet to display pet behaviors while the child performs the training task, and continuously acquiring multiple sets of multimodal physiological and behavioral data of the child during the training task at preset time intervals, each set of the multimodal physiological and behavioral data including eye movement trajectory data and facial images; Step S30: Based on the multimodal physiological and behavioral data of the current group, the child's emotional activation level, gaze stability, and red zone drift index are obtained, and the child's emotional activation level, gaze stability, and red zone drift index are input into the trained decision tree model to obtain the current cognitive load assessment result, which includes the cognitive load state and concentration level; Step S40: Using the current cognitive load assessment result as an updated cognitive load assessment result, and dynamically controlling the task and dynamically adjusting the pet behavior displayed by the virtual pet when the child performs the next training task based on the updated cognitive load assessment result; Step S50: When the training duration reaches the target training duration, the virtual scene is closed, the concentration training is evaluated, and the concentration training result is obtained.

2. The method for training children's concentration according to claim 1, characterized in that: The training method further includes the following steps before step S10: obtaining basic profile information of the child, wherein the basic profile information includes a baseline of attention level, a range of physiological parameters, and basic information of the user; The task scenario loaded in step S10 is determined based on the basic file information.

3. The method for training children's concentration according to claim 1, characterized in that: The step of obtaining the emotional activation degree includes: Step (1), preprocessing the facial images collected within the preset time interval to obtain preprocessed facial images, wherein the preprocessing includes cropping the facial area and image standardization, wherein the number of the facial images is one frame or multiple frames; Step (2), extracting features of each frame of preprocessed facial image respectively, obtaining at least one facial feature set corresponding to each frame of preprocessed facial image, and inputting each facial feature set into an expression classification model respectively, obtaining a probability distribution result of the expression classification corresponding to each frame of facial image, wherein the probability distribution result of the expression classification includes an expression classification name and a probability of the expression classification, and the expression classification name includes happiness, surprise, anger, disgust, sadness, fear and neutrality; Step (3), based on the preset typical emotion activation degree corresponding to each type of expression and the probability distribution result of the expression classification corresponding to each frame of the facial image, obtaining the emotion activation degree corresponding to each frame of the facial image; Step (4): obtaining an average value of the emotion activation levels corresponding to all facial images, and using the average value of the emotion activation levels as the emotion activation level of the child.

4. The method for training children's concentration according to any one of claims 1 to 3, characterized in that: The eye movement trajectory data includes the coordinates of the gaze point, the gaze time, and the scanning path and distance. The step of obtaining the gaze stability includes: Step (1), projecting the coordinates of each gaze point into the interface coordinate system of the task scene, and forming a gaze point cloud map based on the gaze time, scanning path and distance; Step (2) divides the gaze point cloud map into two-dimensional grids, calculates the cumulative gaze time for each grid, performs fuzzy weighting on the point cloud using Gaussian kernel density estimation, and generates a red zone migration map with a gradient of gaze point color; Step (3) is to calculate gaze stability based on the key cognitive sites preset in the task scene and the red zone migration map according to the first formula, wherein the first formula is: Among them, FS is gaze stability, T target is the total duration of fixation on the key cognitive point, T total is the total duration of all fixations.

5. The method for training children's concentration according to claim 4, characterized in that: Based on the eye movement trajectory data, the red zone drift index is calculated according to a second formula, wherein the second formula is: DI is the red zone drift index, (x i ,y i ) is the coordinate of the i-th fixation point, and N is the total number of fixations.

6. The method for training children's concentration according to claim 4, characterized in that: The concentration training results include the cognitive load curve and red zone migration diagram generated in real time during the training process, and the visual concentration trend report output after the training is completed.

7. The method for training children's concentration according to claim 6, characterized in that: The training method also includes storing the training records and concentration training results of each training session, and dynamically adjusting the task type of the next training task, the visual layout of the task scene, and the interaction frequency based on the historical training records and the red zone migration map obtained from the current training session, so as to load appropriate task scenes for children with different cognitive styles.

8. The method for training children's concentration according to claim 1, characterized in that: The pet behaviors include prompting interaction behaviors for guiding or prompting current task goals, rewarding interaction behaviors for providing visual and emotional encouragement after completing a task, and accompanying interaction behaviors for providing emotional comfort and non-task oriented interactions.

9. The method for training children's concentration according to claim 1 or 8, characterized in that: The concentration level includes four levels: high, medium, low and very low. Step S40 also includes comparing the current concentration level with the previous concentration level, and when the concentration level drops directly from high to low, adjusting the virtual pet to display the pet behavior corresponding to the short-term decrease.

10. A children's concentration training system, applied to a wearable device loaded with a virtual pet interactive platform, characterized in that: The training system comprises: A loading module, configured to receive a loading instruction and load a virtual scene for children's concentration training based on the loading instruction, wherein the virtual scene includes a task scene corresponding to the training task and a virtual pet for accompanying the child in training; a first acquisition module, configured to control the virtual pet to display pet behaviors while the child performs the training task, and continuously acquire multiple sets of multimodal physiological and behavioral data of the child during the training task at preset time intervals, each set of the multimodal physiological and behavioral data including eye movement trajectory data and facial images; A second acquisition module is configured to acquire the child's emotional activation level, gaze stability, and red zone drift index based on the multimodal physiological and behavioral data of the current group, and input the child's emotional activation level, gaze stability, and red zone drift index into a trained decision tree model to obtain a current cognitive load assessment result, wherein the cognitive load assessment result includes a cognitive load state and a concentration level; an adjustment module, configured to use the current cognitive load assessment result as an updated cognitive load assessment result, and to dynamically control the task and dynamically adjust the pet behavior displayed by the virtual pet when the child performs the next training task based on the updated cognitive load assessment result; The third acquisition module closes the virtual scene when the training time reaches the target training time, evaluates the concentration training, and obtains the concentration training result.

Images