A progressive training method and device pre-adapted to a real scene

Abstract

The application relates to the technical field of electroencephalogram signal monitoring and data processing, in particular to a pre-adaptation to a real scene progressive training method and device, which comprises the following steps: acquiring scene information of a real scene; performing element level splitting on the scene information; respectively performing progressive level modeling according to elements at different levels after splitting; constructing a multilevel virtual scene simulating real environments at different levels; acquiring heterogeneous feature expression of a user when the user is pre-adapting in the multilevel virtual scene; dynamically switching levels of the multilevel virtual scene according to the heterogeneous feature expression; and performing pre-adaptation to the real scene progressive virtual sensory stimulation training in the multilevel virtual scene at the switched level. By constructing a multilevel virtual scene strongly associated with the real scene, the problem that the existing training is disconnected with the real scene is solved, and by virtue of the multilevel progressive scene version design, different training stages can be adapted, so that the training of the user is more flexible.

Description

Technical Field

[0001] This application relates to the field of EEG signal monitoring and data processing technology, and in particular to a progressive training method and device for pre-adaptation to real-world scenarios. Background Technology

[0002] Some children with autism spectrum disorder (ASD) have sensory processing abnormalities, resulting in a significantly lower threshold for sensory stimulation tolerance than typically developing individuals. They are abnormally sensitive to sensory stimuli such as sound and light. Environments that typically developing individuals can easily adapt to, such as flashing lights or sudden increases in noise, may trigger stress-induced out-of-control behaviors in children with ASD, affecting their safety when going out and their ability to adapt to society.

[0003] Current intervention methods for adapting some ASD children to outdoor scenarios still have problems such as scenarios being detached from real environments, lack of targeted stressor design, and delayed risk warnings. Summary of the Invention

[0004] The purpose of this application is to provide a progressive training method and device that is pre-adapted to real-world scenarios. The progressive, multi-level virtual scenarios are generated based on real-world environmental data, with strong correlation between stressors and scenarios, which significantly improves the realism and adaptability of the virtual scenarios, solves the problem of existing training being disconnected from real-world scenarios, and can also adapt to different training stages through the multi-level progressive scenario version design, making user training more flexible.

[0005] In some embodiments, this application provides a progressive training method for pre-adapting to a real-world scenario. The method includes: acquiring scene information of a real-world scenario; performing element-level segmentation on the scene information; performing progressive hierarchical modeling based on the elements at different levels after segmentation to construct a multi-level virtual scene simulating different levels of real-world environments; acquiring heterogeneous feature expressions of a user during pre-adaptation in the multi-level virtual scene; dynamically switching the levels of the multi-level virtual scene based on the heterogeneous feature expressions; and performing progressive virtual sensory stimulation training for pre-adaptation to a real-world scenario in the multi-level virtual scene after the switching.

[0006] In some embodiments, the step of obtaining the heterogeneous feature expression of the user during pre-adaptation in the multi-level virtual scene and dynamically switching the level of the multi-level virtual scene based on the heterogeneous feature expression includes: obtaining the heterogeneous feature expression of the user during pre-adaptation in the current level virtual scene; determining the user's fitness level in the current level virtual scene based on the sensing parameters and interaction parameters in the heterogeneous feature expression; and dynamically switching the level of the virtual environment in which the user is located based on the fitness level.

[0007] In some embodiments, determining the user's adaptability in the current level of the virtual scene based on the sensing parameters and interaction parameters in the heterogeneous feature representation includes: determining the environmental adaptability corresponding to the user's physiological stability based on the user's sensing parameters and / or interaction parameters; determining the stress response degree corresponding to the user's behavioral performance based on the user's sensing parameters and / or interaction parameters; determining the task completion degree corresponding to the user's task completion quality based on the user's sensing parameters and / or interaction parameters; determining the indicator stability degree corresponding to the user's stability verification based on the user's sensing parameters and / or interaction parameters; determining the care assessment degree corresponding to the user's caregiver collaboration based on the user's sensing parameters and / or interaction parameters; and determining the user's adaptability in the current level of the virtual scene based on at least one of the environmental adaptability, the stress response degree, the task completion degree, the indicator stability degree, and the care assessment degree.

[0008] In some embodiments, determining the user's fitness level in the current level virtual scene based on the sensing parameters and interaction parameters in the heterogeneous feature representation, and determining the user's level adaptability based on the fitness level, includes: if the user's fitness level in the current level virtual scene determined based on the sensing parameters and / or interaction parameters meets the switching conditions, then switching from the current level virtual scene to the next level virtual scene, so that the user can undergo virtual sensory stimulation training to pre-adapt to the real scene in the next level virtual scene, and iteratively adjusting the level of the virtual scene based on the user's fitness level until the user completes pre-adaptation to the real scene in the highest level virtual scene. The training process involves either virtual sensory stimulation in a real-world scenario or termination of training. If the user's adaptation level in the current virtual scenario, as determined by the sensor parameters and / or interaction parameters, does not meet the switching conditions, the system continues to check whether the user has a rollback record that has reached the preset maximum number of rollbacks in the current virtual scenario. If so, a risk warning is issued or training is terminated. If not, the user continues to undergo virtual sensory stimulation training in the current virtual scenario to adapt to the real-world scenario. The system iteratively adjusts the level of the virtual scenario based on the user's adaptation level until the user completes virtual sensory stimulation training in the highest-level virtual scenario to adapt to the real-world scenario or terminates training.

[0009] In some embodiments, the step of hierarchically splitting the scene information into elements and progressively modeling the elements at different levels to construct multi-level virtual scenes simulating different levels of real environments includes: hierarchically splitting the environmental elements and social context elements in the scene information to obtain environmental elements and social context elements at different levels; constructing corresponding levels of virtual scenes using the environmental elements and social context elements at different levels; wherein, the virtual scenes at different levels are used to progressively simulate real scenes, and as progressive training progresses, the virtual scenes continuously approach the real scenes.

[0010] In some embodiments, constructing virtual scenes of corresponding levels using environmental elements and social context elements at different levels includes: when a user is in a key-level virtual scene for the first time to undergo virtual sensory stimulation training to adapt to a real scene, the user's sensory hypersensitivity is not set in the environmental elements of the key level to construct the key-level virtual scene; when a user is not in a key-level virtual scene for the first time to undergo virtual sensory stimulation training to adapt to a real scene, the user's sensory hypersensitivity is dynamically set in the environmental elements of the key level to construct the key-level virtual scene, wherein the key level is a sub-high level.

[0011] In some embodiments, the calculation step of the physiological stability feature interval in the environmental adaptability includes: obtaining a first parameter of the user in the current scenario according to a safe environment standard, the first parameter including an environmental comfort reference value; obtaining a scenario risk correction weight corresponding to the current scenario category; adjusting the first parameter according to a scenario risk correction strategy, a scenario risk correction weight, and a sensitivity correction strategy to obtain environmental critical conditions; obtaining the user's physiological stability features in the current scenario category and current activity state; correcting the physiological stability features using a scenario risk correction strategy to obtain the user's physiological stability feature interval; collecting the user's electroencephalogram (EEG) signal; calculating the user's fitness correction value for the training device based on the EEG signal; and correcting the physiological stability feature interval based on the fitness correction value.

[0012] In some embodiments, the step of collecting the user's electroencephalogram (EEG) signals, calculating the user's fitness correction value for the training device based on the EEG signals, and correcting the physiological stability feature interval based on the fitness correction value includes: collecting the user's resting-state EEG signals when not wearing the training device and calculating a first set of feature parameters; collecting the user's resting-state EEG signals after the child wears the training device and calculating a second set of feature parameters; determining the user's fitness correction value for the training device based on the difference between the first set of feature parameters and the second set of feature parameters; determining a scaling factor based on the fitness correction value; and correcting the physiological stability feature interval based on the scaling factor.

[0013] In some embodiments, the method further includes: if at least one of the physiological stability feature interval, the stress response degree, the task completion degree, the indicator stability degree, and the care assessment degree triggers a high-risk threshold of the corresponding level of virtual environment, the virtual sensory stimulation training pre-adapted to the real scene is terminated.

[0014] In some embodiments, this application also provides a progressive training device for pre-adapting to real-world scenarios. The device includes: a modeling module, configured to acquire scene information of a real-world scenario, perform element-level segmentation of the scene information, and perform progressive hierarchical modeling based on the elements at different levels after segmentation to construct a multi-level virtual scene simulating different levels of real-world environments; a multimodal data synchronous acquisition module, configured to acquire heterogeneous feature expressions when a user wears a training device and performs pre-adaptation in the multi-level virtual scene; a multimodal feature fusion analysis module, configured to calculate the user's sensory parameters and / or interaction parameters based on the heterogeneous feature expressions; and a training interaction module, configured to dynamically switch the levels of the multi-level virtual scene based on the user's sensory parameters and / or interaction parameters, and perform pre-adaptation training in the multi-level virtual scene after the switching.

[0015] The above embodiments provide a progressive training method and apparatus pre-adapted to real-world scenarios. This method is a modeling approach for virtual scenarios based on real-world scenarios, allowing users to conduct virtual sensory stimulation training within a constructed virtual scenario simulating a real-world environment. Unlike existing virtual training methods based on preset virtual scenarios, this application utilizes real-time or near-real-time environmental information of the destination to dynamically construct a highly realistic virtual training environment. It also specifically introduces dynamic stressors common in this environment category that may cause user discomfort, enabling users to conduct progressive virtual sensory stimulation training pre-adapted to real-world scenarios within the constructed virtual scene. Furthermore, the multi-level progressive scenario version design can adapt to different training stages, making user training more flexible. Attached Figure Description

[0016] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, wherein:

[0017] Figure 1 This is a schematic diagram of a progressive training process for pre-adapting to a real-world scenario, provided in one embodiment of this application.

[0018] Figure 2 This is a schematic diagram illustrating different virtual scene stage levels and the switching between stage levels, provided in one embodiment of this application;

[0019] Figure 3This is a flowchart of a personalized threshold adaptation and calibration process provided in one embodiment of this application;

[0020] Figure 4 This is a flowchart of a personalized threshold adaptation and calibration process provided in another embodiment of this application;

[0021] Figure 5 This is a schematic diagram illustrating the process of switching between different levels of a virtual scene, provided in one embodiment of this application.

[0022] Figure 6 This is a flowchart of hierarchical modeling and progressive training provided in one embodiment of this application;

[0023] Figure 7 This is a schematic diagram of a progressive training device module for pre-adapting to real-world scenarios, provided in one embodiment of this application.

[0024] Figure 8 This is a schematic diagram of the structure of a computer device provided in one embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0026] The technical solutions of the various embodiments of this application can be combined with each other, but only if they are based on the ability of a person skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by this application.

[0027] This solution does not aim to obtain disease diagnosis results or health status, but is merely a method for processing information such as scene information and user sensor data, and all steps are information processing methods implemented by computers and other devices.

[0028] It should be fully understood that the user information involved in this application (including but not limited to user physiological information, user personal information, etc.) is information and data authorized by the user or fully authorized by all parties. The use of user information shall comply with the privacy policies and practices of the industry that are generally considered to meet or exceed the requirements for maintaining user privacy. The collection, use and processing of related data shall comply with relevant laws, regulations and standards, and provide corresponding operation access points for users to choose to authorize or refuse.

[0029] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, specific embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0030] See Figure 1 This application provides a progressive training method for pre-adapting to real-world scenarios, comprising:

[0031] Step S101: Obtain scene information of the real scene.

[0032] Step S102: The scene information is split into elements at different levels, and progressive hierarchical modeling is performed according to the elements at different levels after splitting to construct a multi-level virtual scene that simulates different levels of real environment.

[0033] Step S103: Obtain the heterogeneous feature expression of the user during pre-adaptation in the multi-level virtual scene, dynamically switch the level of the multi-level virtual scene according to the heterogeneous feature expression, and conduct progressive virtual sensory stimulation training for pre-adaptation to the real scene in the multi-level virtual scene after the switch.

[0034] In this context, the real-world scene can be the user's destination. Scene information within the real-world scene can include environmental elements and social context elements. Environmental elements can include plants, buildings, animals, etc., while social context elements include the people interacting with the user, including the number and identities of these individuals. In some embodiments, the scene elements used to construct different levels of virtual scenes may differ, specifically including variations in environmental elements and social context elements.

[0035] In some embodiments, the scene information is hierarchically split into elements, and progressive hierarchical modeling is performed on the elements at different levels after splitting to construct multi-level virtual scenes that simulate real environments at different levels. This includes: splitting the environmental elements and social context elements in the scene information into elements at different levels according to their levels to obtain environmental elements and social context elements at different levels; constructing virtual scenes at corresponding levels using the environmental elements and social context elements at different levels; wherein, the virtual scenes at different levels are used to progressively simulate real scenes, and as progressive training progresses, the virtual scenes continuously approach the real scenes.

[0036] In some embodiments, virtual modeling can specifically be VR modeling, XR modeling, and MR modeling, etc., and a multi-level virtual scene can include four levels. The following embodiments use VR modeling and the construction of a four-level virtual scene as an example for illustration. However, it is understood that other embodiments may also include virtual scenes with other numbers of levels, and this application is not limited to a four-level virtual scene.

[0037] This application includes a VR modeling module, which constructs an immersive VR scene containing dynamic stressors and social contexts based on real-time environmental information of the destination in a real-world scenario. It further breaks down scene elements in the real-world scene to obtain multi-level scene elements, and constructs corresponding multi-level progressive virtual scenes based on these multi-level scene elements. Specifically, the multi-level virtual scenes may include a basic virtual scene, a simplified virtual scene, a near-realistic virtual scene, and a fully realistic virtual scene. The generation of these four progressive multi-level virtual scenes meets the user's scenario requirements for progressive training.

[0038] This application also includes a VR training interaction module, which guides users to experience and attempt to cope with preset stimuli and social interactions in simulated environments corresponding to multi-level virtual scenes, and records the user's training performance data in the corresponding level of virtual scenes. Specifically, it can be divided into four progressive multi-level training stages. In the corresponding multi-level virtual scenes, the user's performance is evaluated based on quantitative indicators such as physiological, behavioral, and task-related metrics. The training difficulty is switched by combining the stability of the user's performance with feedback from the guardian, realizing dynamic switching of levels in dynamic multi-level virtual scenes. By integrating a progressive training mode management subsystem, the training difficulty can be dynamically adjusted. It should be noted that the four levels in the above embodiments are only one embodiment, and other embodiments may have other numbers of levels.

[0039] In some embodiments, different levels of virtual scenes are constructed based on different environmental and social context elements. The following example illustrates the construction of four different levels of virtual scenes.

[0040] In the first-level virtual scenario (basic version), the core objective is to eliminate the unfamiliarity of the VR environment and establish an initial sense of security. The training task is to follow a virtual caregiver along a fixed route to familiarize the child with the environment. Only environmental elements are retained in the scene features, such as plants, buildings, and animals. Strangers are not included in the social context elements; a virtual caregiver can be generated solely based on a guardian familiar to the child, and this virtual caregiver is always visible. Furthermore, all environmental factors are set within safe thresholds.

[0041] In the second-level virtual scenario (simplified version), the core objective is to adapt to mild, sudden stimuli, autonomously complete simple tasks, and attempt passive social responses. Training tasks involve autonomously exploring the virtual environment and completing simple tasks such as finding designated targets, like the pink slide in a park scene or the potato chip shelf in a supermarket scene. This includes coping with low-intensity, single-environment stimuli and responding to interactions with virtual pedestrians. Environmental and social elements include randomly added preset numbers of virtual pedestrians (e.g., three to five people), who will proactively initiate friendly interactions with the child. Single environmental disturbances are randomly introduced at specific times, but all environmental signals remain within safe thresholds. The system provides a notification before the stimulus is introduced, and the virtual caregiver is always visible.

[0042] In the third-level virtual scenario (close to reality), the core objective is to adapt to complex stimuli and complete simple proactive social interactions. Training tasks involve autonomously exploring the virtual environment and completing simple tasks such as finding designated targets, like the pink slide in a park scene or the potato chip shelf in a supermarket scene. This involves coping with complex stimuli and common social situations within the scenario to complete proactive social tasks. Environmental and social scenario elements include a completely realistic recreation of real pedestrian traffic, with random introduction of complex environmental interference at specific times. The interference factors are similar to but do not exceed safe thresholds. The system provides no notification when stimuli are introduced, and typical and common social situations within the scenario are set up, with the virtual caregiver always visible.

[0043] In the fourth-level virtual scenario (fully realistic version), the core objective is to adapt to real-world scenarios and cope with potential overload stimuli. Training tasks involve autonomously exploring the virtual environment and completing simple tasks such as finding designated targets, like the pink slide in a park scene or the potato chip shelf in a supermarket scene. This involves coping with real-world stimuli, including unpredictable potential overload stimuli, and handling common social situations within the scenario, completing proactive social tasks. Environmental and social scenario elements include a completely realistic recreation of real-world pedestrian traffic and environmental factors. If any environmental factor exceeds a safety threshold, the system automatically records it and sets typical and common social situations within the scenario. The virtual caregiver does not appear in the child's field of vision, but will be informed before training begins that the caregiver is available at any time.

[0044] The different levels of virtual scenes are shown in Table 1 below.

[0045] Table 1. Virtual Scenes at Different Levels

[0046]

[0047] It should be noted that the basic version is the starting point for virtual training that pre-adapts to real-world scenarios, the simplified version is an initial increase in complexity of the basic version, the fully realistic version is the endpoint of pre-adaptive training, and the near-realistic version is a simplification of the fully realistic version.

[0048] See Figure 2 The arrows indicate the phase transition process. Dashed arrows represent version transitions, and the degree of difficulty escalation is related to the complexity of the actual environment at the destination. Solid arrows represent version transitions, where the difficulty escalation is controlled by expert experience, and the difficulty range is relatively small. Furthermore, there may be rollback iterations during version transitions.

[0049] In some embodiments, this application also addresses single environmental interference and complex environmental interference. For example, sudden music or flashing billboards are each single environmental interference (e.g., sound, light flash), while complex environmental interference is a random combination of single environmental interferences. Different levels of virtual scenes are constructed using different environmental interferences to achieve the construction of progressive virtual scenes.

[0050] In the above embodiments, the introduction of environmental interference at specific time points in stages two and three includes setting the shortest time interval between different environmental interferences as t, and setting the number of times environmental interference is introduced in this training stage as n. Therefore, in this training stage, n time points are randomly sampled and selected, and environmental interference is introduced with the interval between the time points greater than t.

[0051] In the above embodiments, the method of modeling virtual scenes based on real scenes, the multi-level progressive virtual scenes are generated based on real environment data, and the stressors are strongly correlated with the scene, which significantly improves the realism and adaptability of the virtual scenes, solves the problem of existing training being disconnected from real scenes, and the multi-level progressive scene version design can also adapt to different training stages, making the user's training more flexible.

[0052] The above embodiments provide a progressive training method pre-adapted to real-world scenarios. This method of modeling virtual scenarios based on real-world scenarios allows users to conduct virtual sensory stimulation training in a constructed virtual scenario that simulates a real-world scenario. This application's dynamic virtual modeling based on real-world environmental information differs from existing virtual training methods based on preset virtual scenarios. This application utilizes real-time or near-real-time environmental information of the destination, such as images, videos, and audio, to dynamically construct a highly realistic virtual training environment. It also specifically introduces common dynamic stressors (e.g., sudden high-decibel noise, bright flashing lights, etc.) and typical social situations (e.g., shopping, queuing, etc.) that may cause user discomfort within this environmental category. This allows users to conduct progressive virtual sensory stimulation training pre-adapted to real-world scenarios within the constructed virtual scenario.

[0053] In some embodiments, acquiring heterogeneous feature representations of a user during pre-adaptation in the multi-level virtual scene, and dynamically switching the levels of the multi-level virtual scene based on the heterogeneous feature representations, includes: acquiring heterogeneous feature representations of a user during pre-adaptation in the current level of the virtual scene; determining the user's fitness level in the current level of the virtual scene based on sensing parameters and interaction parameters in the heterogeneous feature representations; and dynamically switching the level of the virtual environment in which the user is located based on the fitness level.

[0054] Heterogeneous feature representation refers to the use of multiple data features of different dimensions and types to comprehensively evaluate the user's state and switch the corresponding level of the virtual world.

[0055] Specifically, this application includes a multimodal data synchronous acquisition module that utilizes multimodal sensors to collect the user's physiological signals, behavioral data, and training performance data in real time. The physiological signals may include one or more of the following: electroencephalogram (EEG), electrooculogram (EOG), galvanic skin response (GSR), and heart rate variability (HRV). The behavioral data may include one or more of the following: voice, tone of voice, and behavioral states captured by the camera.

[0056] This application also includes a multimodal feature fusion analysis module, which extracts physiological signal features and behavioral features, obtains task-related training data from the training system, monitors the user's status, evaluates the user's training performance and adaptability, and feeds the evaluation results back to the VR training module to control the switching of training stages, so that the user can switch to a virtual scene level that matches their performance.

[0057] This application also includes an assessment result output and early warning module. This module generates an adaptation assessment report, which may include a phase progress chart and provide preventative suggestions when high-risk discomfort is detected. Specifically, the output may include progress visualization, such as a phase achievement progress bar, like 100% physiological achievement, 80% task achievement, and 0 behavioral abnormalities. Furthermore, the corresponding output can be synchronized to the caregiver's backend. If the assessed user fails to successfully complete all training phases, a destination risk warning can be further pushed to the guardian.

[0058] In some embodiments, determining a user's adaptability in the current level of the virtual scene based on the sensing parameters and interaction parameters in the heterogeneous feature representation includes: determining an environmental adaptability corresponding to the user's physiological stability based on the user's sensing parameters and / or interaction parameters; determining a stress response corresponding to the user's behavioral performance based on the user's sensing parameters and / or interaction parameters; determining a task completion rate corresponding to the user's task completion quality based on the user's sensing parameters and / or interaction parameters; determining an indicator stability rate corresponding to the user's stability verification based on the user's sensing parameters and / or interaction parameters; determining a care assessment rate corresponding to the user's caregiver collaboration based on the user's sensing parameters and / or interaction parameters; and determining the user's adaptability in the current level of the virtual scene based on at least one of the environmental adaptability, the stress response rate, the task completion rate, the indicator stability rate, and the care assessment rate.

[0059] In some embodiments, heterogeneous feature representation includes sensing parameters and interaction parameters. Furthermore, the sensing parameters can be used to assess physiological stability and behavioral performance, while the interaction parameters include task completion quality, stability verification, and caregiver collaboration confirmation.

[0060] In some embodiments, the user's environmental fit can be determined based on the proportion of a target frequency band in the user's EEG data and the relationship between the target physiological signal and the physiological stable feature interval. The target frequency band in the EEG data can be the alpha wave in the acquired EEG, and the target physiological signal can be physiological signals such as GSR and HRV. Specifically, the user's environmental fit can be determined based on the proportion of alpha waves in the EEG and the relationship between physiological signals such as GSR and HRV and the physiological stable feature interval.

[0061] In some embodiments, the user's stress response level can be determined based on the frequency of the user's avoidance behavior and the frequency of the user's stress response behavior. Specifically, during user training, the frequency of the user's avoidance behavior and the frequency of the user's stress response behavior can be determined using data collected by cameras and sensors, thereby determining the user's stress response level.

[0062] In some embodiments, a user's task completion rate can be determined based on the user's task completion rate, the number of task interruptions, and the task completion time. Specifically, during the user's training process, the user's task completion rate, the number of task interruptions, and the task completion time can be determined using data collected through VR training interaction modules, etc., thereby determining the user's task completion rate.

[0063] In some embodiments, the stability of a user's metrics can be determined based on the number of times the user continuously meets the environmental adaptability, stress response, and task completion criteria for completing a certain stage of training. Specifically, the stability of a user's metrics can be determined during the user's training process through a multimodal fusion analysis module.

[0064] In some embodiments, the user's care assessment level can be determined based on the caregiver's feedback.

[0065] Further, the user's hierarchical adaptability is determined based on at least one of the environmental adaptability, stress response, task completion, indicator stability, and care assessment. The hierarchical level of the virtual environment in which the user resides is dynamically switched based on the hierarchical adaptability. In some embodiments, the user's adaptability in the current level of the virtual scene is determined based on the sensing parameters and interaction parameters in the heterogeneous feature representation. Determining the user's hierarchical adaptability based on the adaptability includes: if the user's adaptability in the current level of the virtual scene, determined based on the sensing parameters and / or interaction parameters, meets the switching conditions, then switching from the current level of the virtual scene to the next level of the virtual scene is executed, so that the user can undergo virtual sensory stimulation training to pre-adapt to the real scene in the next level of the virtual scene. The hierarchical level of the virtual scene is iteratively adjusted based on the user's adaptability until the user completes pre-adaptation to the real scene in the highest level of the virtual scene. The training process involves either virtual sensory stimulation in a real-world scenario or termination of training. If the user's adaptation level in the current virtual scenario, as determined by the sensor parameters and / or interaction parameters, does not meet the switching conditions, the system continues to check whether the user has a rollback record that has reached the preset limit in the current virtual scenario. If so, a risk warning is issued or training is terminated. If not, the user continues to undergo virtual sensory stimulation training in the current virtual scenario to adapt to the real-world scenario. The system iteratively adjusts the level of the virtual scenario based on the user's adaptation level until the user completes virtual sensory stimulation training in the highest virtual scenario to adapt to the real-world scenario or terminates training.

[0066] See Figure 5This document provides a flowchart illustrating the process of switching between different scene levels. First, it determines whether the required number of stability verification iterations for the corresponding level has been met. If not, training continues at the current level. If the required number of stability verification iterations has been met, it further determines whether the conditions for switching to stage n+1 are met. If so, training switches to stage n+1. If not, it further determines whether there are any rollback records in the current stage n. If there are no rollback records, training continues at the current level; otherwise, training ends, and a destination risk warning is sent to the monitoring personnel. In some embodiments, the maximum number of rollbacks a user can perform in the current stage is one. It is understood that in other embodiments, the maximum number of rollbacks a user can perform can be other numbers, not limited to one, and can be flexibly set according to actual needs. In other embodiments, the maximum number of rollbacks a user can perform in different virtual environments may vary. For example, as the difficulty of virtual training increases, the maximum number of rollbacks a user can perform also increases accordingly.

[0067] In some embodiments, switching conditions adapted to different levels of virtual scenarios are pre-set. Specifically, to ensure the objectivity, stability, and individual adaptability of switching during the training phase, the switching conditions can be determined based on multimodal quantitative data, such as physiological signals, behavioral performance, task completion quality, stability of continuous training performance, and caregiver collaboration confirmation, forming a three-layer judgment logic of multi-dimensional indicator achievement, stability verification, and safety backup. Multi-dimensional achievement means simultaneously meeting the three dimensions of physiological stability, behavioral performance, and task completion quality to avoid misjudgment based on a single indicator. Stability verification requires continuous achievement, such as two or three consecutive training sessions, rather than a single accidental achievement, to ensure the user's ability is consistent. Caregiver collaboration confirmation means that the caregiver, through background observation, confirms that the user remains calm and does not express resistance during training, such as not wanting to play or being afraid.

[0068] In the following embodiments, a virtual scene comprising four levels is used as an example for illustration.

[0069] The transition from Phase 1 to Phase 2 (from the first-level virtual scene to the second-level virtual scene) aims to confirm that the user has adapted to the VR environment, exhibits no basic stress response, and can complete simple command tasks. The assessment dimensions include physiological stability, behavioral performance, task completion quality, stability verification, and caregiver collaboration confirmation. Physiological stability involves acquiring the user's EEG, GSR, HRV, and other physiological signals from a multimodal data acquisition module and further evaluating whether the acquired signals meet the required indicators. For example, whether the proportion of alpha waves in the EEG is greater than or equal to r1 (alpha waves reflect a relaxed state, and a higher proportion indicates greater stability), and whether GSR, HRV, and other physiological signals are all within the physiological stability characteristic range. Behavioral performance includes obtaining the user's avoidance behavior data or stress behavior data from camera or sensor behavior recognition and further evaluating whether the acquired data meets the required indicators. For example, whether the frequency of avoidance behavior is zero in each training session (avoidance behaviors include covering eyes, backing away, crying, etc.), and whether the frequency of strong stress response behavior is zero in each training session (strong stress response behaviors include screaming, curling up, etc.). The evaluation criteria for task completion include: 1) whether the user's task completion rate reaches 100% as determined by the VR training interaction module; 2) whether the user completes the follow-along walking exercise; 3) whether the user's task interruption frequency is zero; 4) whether the user's average completion time is less than the preset time (e.g., 5 minutes); and 5) whether the user exhibits significant procrastination. Stability verification includes whether the user meets all the above indicators in two consecutive training sessions, as determined by the multimodal fusion analysis module. Caregiver collaboration confirmation includes confirming, through multi-caregiver interaction backend observation, whether the user's emotions remain calm during training and whether there are any expressions of resistance (such as not wanting to play or fear).

[0070] The switching conditions from Phase 1 to Phase 2 are shown in Table 2 below.

[0071] Table 2 Switching conditions from Phase 1 to Phase 2

[0072]

[0073] The transition from Phase 2 to Phase 3 (from the second-level virtual scenario to the third-level virtual scenario) aims to confirm that the user can tolerate low-intensity stressors and complete passive social responses. The assessment dimensions include physiological stability, behavioral performance, task completion quality, stability verification, and caregiver collaboration confirmation. The physiological stability dimension involves acquiring the user's EEG, GSR, HRV, and other physiological signals from a multimodal data acquisition module and further evaluating whether the acquired signals meet the required indicators. For example, whether the proportion of alpha waves in the EEG is greater than or equal to r1. Alpha waves reflect a relaxed state; a higher proportion indicates greater stability, allowing for slight fluctuations and reflecting tolerance to low stimuli. Even if GSR, HRV, and other physiological signals fluctuate, they should still remain within the physiological stability characteristic range. The behavioral performance evaluation includes data on user avoidance or stress response behaviors obtained from camera or sensor behavior recognition and VR interaction data. Further assessment is needed to determine if the collected data meets the required metrics. For example, the frequency of avoidance behaviors should be zero in each training session. Avoidance behaviors can include covering eyes, backing away, crying, and other avoidance resistance behaviors. Particular attention is paid to stimulus warning compliance, ensuring the user maintains their current action and does not flee after seeing the stimulus warning. The frequency of strong stress response behaviors should also be zero in each training session. Strong stress response behaviors can include screaming, curling up, and other strong stress behaviors. The task completion quality evaluation includes whether the user's task completion rate reaches 100%, as determined by camera or sensor behavior recognition and voice acquisition from the VR training interaction module. This includes completing simple tasks such as finding a designated target, passive social responses (e.g., a virtual pedestrian waving), and responses such as nodding, waving, or a soft "hi." The evaluation also includes whether the user's task interruption count is zero, the average completion time is less than a preset time (e.g., 8 minutes), and whether the user exhibits no significant procrastination. The stability verification includes whether the user meets all the above metrics in two consecutive training sessions, as determined by the multimodal fusion analysis module. Among them, caregiver collaboration confirmation includes confirming whether the user's emotions are calm and whether there is any resistance expression (such as not wanting to play or being afraid) during the training process based on the backend observation of multiple caregivers interaction.

[0074] The switching conditions from Phase 2 to Phase 3 are detailed in Table 3 below.

[0075] Table 3 Switching conditions from Phase 2 to Phase 3

[0076]

[0077] The transition from Stage 3 to Stage 4 (from the third-level virtual scenario to the fourth-level virtual scenario) aims to confirm that the child can adapt to complex stimuli and complete designated tasks in common social situations. The assessment dimensions include physiological stability, behavioral performance, task completion quality, stability verification, and caregiver collaboration. Physiological stability involves acquiring the user's EEG, GSR, HRV, and other physiological signals using a multimodal data acquisition module, and further evaluating whether the acquired signals meet the required criteria. For example, the proportion of alpha waves in the EEG should be greater than or equal to r², where alpha waves reflect a relaxed state; a higher proportion indicates greater stability, and moderate fluctuations are permissible, reflecting tolerance to complex stimuli. Even if GSR, HRV, and other physiological signals fluctuate, they should remain within the physiologically stable characteristic range. Behavioral performance includes obtaining the user's avoidance behavior data or stress response data based on camera or sensor behavior recognition and VR interaction data, and further evaluating whether the acquired data meets the required criteria. For example, the frequency of avoidance behavior should be zero times in each training session. Avoidance behavior can include covering eyes, backing away, crying, and other avoidance and resistance behaviors. The frequency of strong stress response behaviors must be zero in each training session. Strong stress response behaviors can include screaming, curling up, and other intense stress behaviors, with particular attention paid to behavioral performance after compound stimuli. Task completion quality includes whether the user's task completion rate reaches 90%, as determined by the VR training interaction module's camera or sensor behavior recognition and voice acquisition. This includes completing simple tasks such as finding a designated target, completing designated tasks in classic social scenarios, such as buying candy and queuing at the supermarket checkout, or saying thank you after receiving a free toy. The number of user task interruptions must be zero, the user's average completion time must be less than the preset time (e.g., 10 minutes), and the user must not exhibit significant procrastination. Stability verification includes whether the user meets all the above indicators in three consecutive training sessions, as determined by the multimodal fusion analysis module. Caregiver collaboration confirmation includes confirming whether the user's emotions are calm and whether there are any expressions of resistance (such as not wanting to play or fear) during training, based on observations of multi-caregiver interaction in the background.

[0078] The switching conditions from stage 3 to stage 4 are detailed in Table 4 below.

[0079] Table 4 Switching conditions from Phase 3 to Phase 4

[0080]

[0081] The transition from Phase 4 to the end of training achievement (the transition from the fourth-level virtual scenario to the end of training achievement) involves several evaluation dimensions, including physiological stability, behavioral performance, task completion quality, stability verification, and caregiver collaboration confirmation. Physiological stability involves acquiring the user's EEG, GSR, and HRV signals from a multimodal data acquisition module and further evaluating whether the acquired signals meet the required criteria. For example, the proportion of alpha waves in the EEG should be greater than or equal to r², where alpha waves reflect a relaxed state; a higher proportion indicates greater stability, with moderate fluctuations allowed, reflecting tolerance to complex stimuli. Even if GSR and HRV signals fluctuate, they should remain within the physiologically stable characteristic range. Behavioral performance involves obtaining the user's avoidance or stress behavior data based on camera or sensor behavior recognition and VR interaction data, and further evaluating whether the acquired data meets the required criteria. For example, the frequency of avoidance behavior should be zero times in each training session. Avoidance behavior can include behaviors such as covering eyes, backing away, crying, and other avoidance and resistance behaviors. The frequency of strong stress response behaviors must be zero in each training session. Strong stress response behaviors can include screaming, curling up, and other intense stress behaviors, with particular attention paid to behavioral performance when environmental stimuli approach or exceed a threshold. Task completion quality includes whether the user's task completion rate reaches 80%, as determined by the VR training interaction module's camera or sensor behavior recognition and voice acquisition. This includes completing simple tasks such as finding a designated target, completing designated tasks in classic social scenarios, such as buying candy and queuing at the supermarket checkout, or saying thank you after receiving a free toy. The user's task interruption count should be less than or equal to one, the user's average completion time should be less than the preset time (e.g., 12 minutes), and the user should not exhibit significant procrastination. Stability verification includes whether the user meets all the above indicators in three consecutive training sessions, as determined by the multimodal fusion analysis module. Caregiver collaboration confirmation includes confirming whether the user's emotions are calm and whether there are any expressions of resistance (such as not wanting to play or fear) during training, based on observations of multi-caregiver interaction in the backend.

[0082] The switching conditions from Phase 4 to the end of training are detailed in Table 5 below.

[0083] Table 5 Switching conditions from Phase 4 to the end of training objectives

[0084]

[0085] With the above-mentioned refined conditions, the stage switching no longer relies on subjective judgment, but is based on quantifiable and verifiable multi-dimensional data to ensure that the training difficulty matches the user's training adaptability and maximize the pre-adaptive training effect.

[0086] It should be noted that regarding the switching conditions for multi-level virtual scene training stages, the control over stage switching becomes increasingly stringent during stability verification, from stage 1 to stage 4. For example, stages 1 and 2 are relatively simple, so switching from stage 1 to stage 2, and from stage 2 to stage 3, only requires meeting the target twice consecutively in the current training stage. Stages 3 and 4 are closer to the real environment and are more difficult, so switching from stage 3 to stage 4, and from stage 4 until the end of training is indicated, requires meeting the target three times consecutively in the current training stage.

[0087] In the above embodiments, the steps for obtaining the alpha wave proportion threshold in the stage switching conditions include: enrolling special users who meet certain diagnostic criteria through a multicenter cross-sectional clinical study (which can be stratified according to the severity of ASD symptoms and age, with ≥30 cases in each subgroup). Collecting resting EEG signals with eyes open in a low-noise, soft-light environment. Collecting EEG signals during cognitive training. For different subgroups, calculating the alpha wave proportion under the two acquisition conditions, and determining the 95% reference range (2.5%-97.5% quantile) of the alpha wave proportion in the relaxed and task-aware states for each subgroup through reliability testing. When the user uses the device, determining the alpha wave proportion threshold from the corresponding subgroup based on their clinical diagnosis score and age. r1 is the lower limit of the alpha wave proportion in the relaxed state for the subgroup, and r2 is the lower limit of the alpha wave proportion in the task-aware state for the subgroup.

[0088] In some embodiments, the method further includes: terminating the virtual sensory stimulation training pre-adapted to the real-world scenario if at least one of the physiological stability characteristic interval, the stress response degree, the task completion degree, the indicator stability degree, and the care assessment degree triggers a high-risk threshold in the corresponding level of the virtual environment. Specifically, this application also provides switching adjustment rules for special situations, including a safety-backup pause mechanism. For example, if any of the following situations occur during training in a virtual scenario, training is immediately paused: when physiological signals exceed the stable interval for 30 seconds, when electromyographic disturbances in EEG increase sharply for more than 10 seconds, or when abnormal behavior occurs, such as loud crying, forcibly removing the device, curling up and refusing to interact. It should be noted that the above time is only for illustrative purposes; in other embodiments, the time value can be customized according to requirements.

[0089] Another embodiment also includes a rollback mechanism. If the switching conditions from training phase n to training phase n+1 are not met, the system automatically rolls back to phase n and retrains for phase n. If, after retraining, it is still impossible to enter the next phase (n+1), then training is paused and terminated.

[0090] In some embodiments, the method provided in this application further includes a real-time monitoring module, which generates a progress bar for each stage of achievement every 2 minutes during the training process through a multimodal fusion analysis module, such as 100% physiological achievement, 80% task achievement, and 0 behavioral abnormalities, and synchronizes it to the caregiver's backend.

[0091] In some embodiments, this application also includes a compliance determination module. After a single virtual scene training session ends, the system automatically compares the detailed indicator requirements of the stage switching and generates a single-session compliance report, such as whether the training session has met the requirements. If there are any non-compliant items, such as the physiological indicator GSR exceeding the stable range, they will be displayed.

[0092] This application also includes a continuous verification module, and the switching conditions for the continuous verification module are different for different stages. For example, for stages 1 and 2, when two consecutive training sessions show that the target has been met, the system triggers an advancement prompt, which the caregiver confirms before switching stages, such as clicking "confirm" to advance to stage 3. For stages 3 and 4, when three consecutive training sessions show that the target has been met, the system triggers an advancement prompt, which the caregiver confirms before switching stages, such as clicking "confirm" to advance to stage 4.

[0093] This application also includes a scene synchronization module. After switching the level of the virtual scene, the VR modeling module can automatically load the virtual scene version corresponding to the next stage. For example, stage 3 corresponds to a near-realistic scene, and the training interaction module updates the tasks and social requirements.

[0094] In another embodiment, for users who consistently fail to meet the target, the system's difficulty level is also downgraded. When a user inputs their destination, the system automatically queries the training logs for the same destination. If more than three "stage rollback, training pause" records exist, a difficulty downgrade correction factor is calculated. ,by The attenuation factor corrects for the intensity of environmental disturbances introduced in all training phases.

[0095] The calculation formula is as follows:

[0096]

[0097] in This represents the lower limit of the proportion of alpha waves in the user's task awareness state corresponding to their subgroup. It is the proportion of the alpha band in the resting-state EEG signal when the user is not wearing a VR device.

[0098] In some embodiments, constructing virtual scenes of corresponding levels using environmental elements and social context elements at different levels includes:

[0099] When a user is in a key-level virtual scene for the first time for pre-adaptation to a real-world virtual sensory stimulation training, no sensory hypersensitivity items are set for the user in the key-level environmental elements during the construction of the key-level virtual scene; when a user is not in a key-level virtual scene for the first time for pre-adaptation to a real-world virtual sensory stimulation training, sensory hypersensitivity items are dynamically set for the user in the key-level environmental elements during the construction of the key-level virtual scene, wherein the key level is the next higher level.

[0100] In some embodiments, this application also sets user sensory hypersensitivity items in key levels of the virtual scene. For example, complex environmental interference can be set in stages 3 and 4. Before switching from stage 3 to stage 4, three consecutive training sessions in stage 3 are required to meet the target. For example, complex environmental interference is introduced N times in stage 3. In the first training session of stage 3, the user's sensory hypersensitivity items must be avoided each time complex environmental interference is introduced. In the second training session of stage 3, the user's sensory hypersensitivity items must be included when N / 2 times complex environmental interference is introduced. In the third training session of stage 3, the user's sensory hypersensitivity items must be included each time complex environmental interference is introduced. For example, based on the user's clinical information, such as a sensory processing ability profiling scale, it can be known that the user is highly sensitive to light flash. Therefore, in the first training session of stage 3, light flash must be excluded when introducing complex stimuli. In stage 4, when the real environment is reproduced at a 1:1 scale, if the user's sensory hypersensitivity items are detected, they are not filtered out but automatically recorded by the system for subsequent review and analysis. By combining the user's sensory hypersensitivity clinical information with the environmental interference introduced, users can gradually adapt to the environment, while also ensuring the comprehensiveness of pre-training.

[0101] In some embodiments, the calculation step of the physiological stability feature interval in the environmental adaptability includes: obtaining a first parameter of the user in the current scenario according to a safe environment standard, the first parameter including an environmental comfort reference value; obtaining a scenario risk correction weight corresponding to the current scenario category; adjusting the first parameter according to a scenario risk correction strategy, a scenario risk correction weight, and a sensitivity correction strategy to obtain environmental critical conditions; obtaining the user's physiological stability features in the current scenario category and current activity state; correcting the physiological stability features using a scenario risk correction strategy to obtain the user's physiological stability feature interval; collecting the user's electroencephalogram (EEG) signal; calculating the user's fitness correction value for the training device based on the EEG signal; and correcting the physiological stability feature interval based on the fitness correction value.

[0102] In some embodiments, this application further includes a fatigue monitoring module. Between phase transitions, there is a rest period of a certain duration during which the user's resting EEG signals are collected, and a fatigue index is calculated. If the fatigue index exceeds a threshold, the system prompts a rest and automatically pauses training. The fatigue index is calculated based on EEG signals from frontal lobe electrodes, and the formula for calculating the fatigue index is as follows:

[0103]

[0104] in, Indicates fatigue index, express relative power ratio of the bands express Band relative power ratio. Generally, a higher ratio indicates a more relaxed state, while a lower ratio reflects higher cognitive activity or a state of stress.

[0105] In some embodiments, the device also includes secondary device adaptation correction. Stages 1 and 2 are of low difficulty and are mainly aimed at familiarizing users with the VR environment and adapting to mild stimuli. When switching from stage 3 to stage 4, the user's adaptation correction value for the entire device will be recalculated based on the resting signal, and the stable physiological characteristic range will be recalibrated.

[0106] In some embodiments, the step of collecting the user's EEG signals, calculating the user's fitness correction value for the training device based on the EEG signals, and correcting the physiological stability feature interval based on the fitness correction value includes: collecting the user's resting-state EEG signals when not wearing the training device and calculating a first set of feature parameters; collecting the user's resting-state EEG signals after the user (child) wears the training device and calculating a second set of feature parameters; determining the user's fitness correction value for the training device based on the difference between the first and second set of feature parameters; determining a scaling factor based on the fitness correction value; and correcting the physiological stability feature interval based on the scaling factor. The first set of feature parameters may include the relative power percentage of the alpha band calculated based on the user's resting-state EEG signals when not wearing the VR device, and may also include other EEG indicators such as the fractal dimension of the EEG. The second set of feature parameters may include the relative power percentage of the alpha band calculated based on the user's resting-state EEG signals after wearing the VR device, and may also include other EEG indicators such as the fractal dimension of the EEG.

[0107] Specifically, the steps to obtain the user's adaptation correction value for the VR device are as follows:

[0108] Collect resting-state EEG signals from users without VR devices and calculate the relative power proportion of the alpha band. (First characteristic parameter group) Collect resting-state EEG signals of users wearing VR devices and calculate the relative power proportion of the α band. (Second characteristic parameter set), thus obtaining the user's initial fitness correction value for the VR device. and with This is a scaling factor used to correct the upper limit of the user's stable physiological characteristic range.

[0109]

[0110] An initial check is performed on various environmental elements of the destination. If more than a certain percentage (e.g., 50%) of the environmental elements exceed the user's comfort threshold, an early warning is issued, reminding the guardian that "the destination is high-risk and not suitable for travel," and no further progressive pre-training is conducted.

[0111] In some embodiments, this application also provides a description of whether a user is suitable for the training system, which is determined based on the resting EEG signals collected by the personalized threshold adaptation module and compared. The lower bound of r1 corresponding to the user's subgroup, if If the system prompts "The user's adaptability to this VR training system is low", it will display the message "The user's adaptability to this VR training system is low".

[0112] Specifically, see Figure 3 as well as Figure 4 This is a personalized threshold adaptation and calibration flowchart. First, the guardian inputs destination information. The system determines the scenario category to which the destination belongs and obtains the scenario risk value for the child from the child's historical behavioral information. Combined with the child's clinical information, it calculates the comfort thresholds for various environmental information within that scenario category. Further, it calculates the physiological stability characteristic range corresponding to the child's physiological signals within that scenario category, also based on the child's clinical information. Furthermore, it includes a preliminary check of various environmental elements of the destination based on the comfort environment thresholds adapted to the child, making a preliminary judgment on the destination's suitability.

[0113] In some embodiments, the physiological stability characteristic interval represents the threshold range within which a user maintains a stable physiological state. Specifically, based on the user's destination information, the scenario category is automatically determined. Combining the user's historical behavioral data and clinical information, the scenario risk value for that scenario is obtained. The comfort environment threshold and stable physiological characteristic interval for that scenario category are calculated. Furthermore, the adaptation correction value for the entire VR device after the user wears it is obtained, thereby correcting the stable physiological characteristic interval under VR device wearing conditions. The steps for determining the user's physiological stability characteristic interval are as follows.

[0114] In some embodiments, the wearable device retrieves the user's clinical information from the memory via a data processor. Specifically, the data processor retrieves the corresponding physiological stability features from the memory based on the user's current activity state. Specifically, the user's current activity state can be determined based on physiological signals and pose information, and the physiological stability feature range corresponding to the user's physiological signals can be determined based on the scene category, the current activity state, and the scene risk feature information corresponding to the current scene category.

[0115] In some embodiments, the step of determining the physiological stability feature interval includes: obtaining a third parameter from the user's historical behavior information, the third parameter including the user's physiological stability features in the current scenario and current activity state; obtaining the physiological stability feature interval based on a third co-correlation relationship reflected by the third parameter, the scenario risk correction strategy, and the scenario risk correction weight; wherein, in the third co-correlation relationship, the physiological stability feature interval increases in response to an increase in the third parameter, the scenario risk correction strategy, and the scenario risk correction weight. In some embodiments, the threshold discrimination and output decision module includes a physiological stability feature interval calculation unit. Specifically, the physiological stability features of the user in the current activity state are obtained from historical behavior information. Based on physiological homeostasis Scenario risk correction strategy Scenario risk correction weight Determine the stable physiological characteristic intervals of each physiological signal. Specifically, calculate the current scene category. Current activity status Below, the stable range of each physiological signal, based on physiological signals For example, its physiological stable characteristic range for:

[0116]

[0117] Among them, the current activity status Next, physiological signals Its stable characteristics include its mean and standard deviation , It is the scenario risk correction weight.

[0118] In the above embodiments, the physiological stability feature interval is obtained through a third synergistic positive correlation. This relationship emphasizes the synergistic effect between multiple factors or variables, jointly promoting the achievement of a certain goal or result. The physiological stability feature interval constructed through this relationship takes into account the correlation and synergy between data from multiple dimensions, making the construction of the physiological stability feature interval more accurate and comprehensive.

[0119] In some embodiments, the scenario category can be determined based on environmental information, and scenario risk characteristic information corresponding to the scenario category can be further obtained. For example, scenario risk characteristic information corresponding to the current scenario category can be obtained from the user's historical behavior information, and a scenario risk correction strategy can be obtained based on the first positive correlation reflected by the scenario risk characteristic information of the scenario category.

[0120] In some embodiments, this application may further include a personal information acquisition module, which includes a historical behavior information acquisition unit to acquire scenario risk characteristic information corresponding to historical scenarios that match the current scenario category.

[0121] Specifically, the wearable device can store the user's historical behavioral data and clinical information in its memory. The data processor in the wearable device can then retrieve the corresponding scenario risk characteristic information from the memory based on the current scenario category.

[0122] In some embodiments, scenario risk characteristic information can also be obtained through large-scale questionnaire surveys of children with ASD, with initial values ​​set based on the obtained baseline population data. Before device startup, scenario risk characteristic information can be personalized based on user history behavior questionnaires completed by guardians. During subsequent use, the device will adjust scenario risk characteristic information based on the number of warnings, alerts, and interventions in each scenario. Guardians can also edit usage logs regarding correct detection, false alarms, and missed alarms through the app's feedback module, thereby helping the device update scenario risk characteristic information. It should be noted that users exhibit unique behavioral changes under scenario overload conditions, and these changes directly affect their scenario risk characteristic information. For example, noisy shopping malls, brightly lit classrooms, and family living rooms have different scenario risk characteristic information.

[0123] In the above embodiments, the scenario risk correction strategy is constructed based on scenario risk characteristic information and the first synergistic positive correlation reflected by the scenario risk characteristic information. A synergistic positive correlation indicates a mutually reinforcing and co-growing relationship between two or more variables; when one variable increases, one or more related variables will also increase accordingly. This relationship emphasizes the synergistic effect between multiple factors or variables, jointly promoting the achievement of a certain goal or result. The scenario risk correction strategy constructed through this relationship takes into account the correlation and synergy between data from multiple dimensions, making the construction of the scenario risk correction strategy more accurate.

[0124] In some embodiments, obtaining a scenario risk correction strategy based on a first positive correlation reflected by scenario risk feature information of scenario categories includes: obtaining scenario risk feature information corresponding to the current scenario category from the user's historical behavior information; obtaining a scenario risk correction strategy based on a first positive correlation reflected by scenario risk feature information of scenario categories; wherein, in the first positive correlation, the scenario risk correction strategy increases in response to an increase in scenario risk feature information.

[0125] The calculation formula for the scenario risk correction strategy is as follows.

[0126]

[0127] in, This is the scenario risk characteristic information for the current scenario category. It is a scenario risk correction strategy.

[0128] In the above embodiments, the scene risk characteristic information is associated with specific target users. This allows for the acquisition of scene risk characteristic information that perfectly corresponds to each user, ensuring that the user's sensory stimulation state can be assessed based on personalized information, resulting in more accurate assessment results. Furthermore, a scene risk correction strategy is derived through a first synergistic positive correlation relationship. This relationship emphasizes the synergistic effect between multiple factors or variables, jointly driving the achievement of a certain goal or result. The scene risk correction strategy constructed through this relationship considers the correlation and synergy between data from multiple dimensions, making the construction of the scene risk correction strategy more accurate.

[0129] Currently, while various intervention methods, including cognitive behavioral training, have been introduced for children with ASD (Autism Spectrum Disorder) to adapt to outdoor scenarios, their practical application still suffers from several significant limitations. For example, the training scenarios are weakly correlated with the real environment. Existing interventions often use pre-set or generic environments, lacking a dynamic connection to the child's actual destination. This leads to a disconnect between training content and real-world scenarios, hindering effective adaptive transfer. Furthermore, there is a lack of individualized adaptation mechanisms. Personalized stimulus configurations and dynamic adjustments are not made based on the hypersensitivity of ASD children (such as abnormal sensitivity to specific sounds and lights), resulting in insufficient targeting and adaptability of training content, making it difficult to accurately simulate the challenges individuals may face in real-world scenarios. Finally, the training process lacks a quantitatively driven, gradual logic. Existing methods often rely on mechanical content changes or fixed-duration training phases, failing to construct a continuous, quantifiable evaluation system based on multi-dimensional data such as physiological indicators and behavioral performance. This results in a lack of dynamic matching between training difficulty and the child's real-time adaptability, making it difficult to achieve truly "gradual" improvement in adaptability.

[0130] Therefore, existing intervention methods are insufficient in terms of scenario realism, individual suitability, and scientific rigor, thus limiting the effectiveness and safety of outdoor adaptation training for children with ASD. There is an urgent need for a systematic approach that can model real-world scenarios, integrate individual characteristics, and dynamically adjust the training process based on objective data to improve the effectiveness of pre-adaptation training for children with ASD before going out.

[0131] This application presents a progressive training method pre-adapted to real-world scenarios. This method, based on VR technology, is a system and approach for assessing and training children with ASD to adapt to their destinations. Its core objective is to construct a training platform highly consistent with actual travel scenarios through multi-version VR scene modeling driven by real-world environments. Through progressive training driven by multimodal data, the training difficulty is dynamically matched to the child's adaptability. Quantitative assessment and risk warnings identify potential discomfort risks in advance and provide a basis for intervention. During training, not only is a VR device adaptation correction step incorporated to establish an adaptation benchmark between the device and the child's physiological stability, but also, combined with clinical information, differentiated intensity adjustments are implemented for hypersensitive elements such as sound, light, and touch, achieving personalized adaptation of training stimuli. Ultimately, this progressively improves the adaptability and social participation of children with ASD in specific real-world travel environments, reduces stress responses in real-world scenarios, and ensures safe travel.

[0132] In one specific embodiment, see Figure 6 This is a flowchart illustrating hierarchical modeling and progressive training provided in one embodiment of this application. This application provides a progressive training method for pre-adapting to real-world scenarios. This method is a multi-level progressive virtual scene modeling process based on real-world scene modeling, and its process includes the following steps.

[0133] 1) Conduct two phase 1 basic adaptation training sessions in the basic training scenario;

[0134] 2) Determine whether the switching conditions from stage 1 to stage 2 are met. If they are met, proceed to step 3); otherwise, proceed to step 14.

[0135] 3) Perform two phase 2 low-intensity training sessions in a simplified scenario;

[0136] 4) Determine whether the switching conditions from stage 2 to stage 3 are met. If they are met, proceed to step 6); otherwise, proceed to step 5.

[0137] 5) Determine if there is a rollback record in stage 2. If there is, proceed to step 14); otherwise, proceed to step 3.

[0138] 6) Obtain the secondary correction value of the user's device fitness and recalibrate the upper limit of the physiological stability characteristic range. (Secondary correction value for fitness) The calculation formula is as follows:

[0139]

[0140] in It refers to the relative power ratio of the α band when the user is wearing VR devices and in a resting state during the transition from stage 2 to stage 3. It is the relative power ratio of the alpha band when the user is not wearing VR devices and is in a resting state, which is obtained in the personalized threshold adaptation step;

[0141] 7) Conduct three phase 3 medium-intensity training sessions in a near-display-quality scene;

[0142] 8) Determine whether the switching conditions from stage 3 to stage 4 are met. If they are met, proceed to step 10); otherwise, proceed to step 9.

[0143] 9) Determine if there is a rollback record in stage 3. If there is, proceed to step 14); otherwise, proceed to step 7.

[0144] 10) Conduct three phase 4 high-intensity training sessions in a completely realistic scenario;

[0145] 11) Determine whether the conditions of "training achieved and training ended" are met. If they are met, proceed to step 13); otherwise, proceed to step 12.

[0146] 12) Determine if there is a rollback record in stage 4. If there is, proceed to step 14); otherwise, proceed to step 10.

[0147] 13) Upon completion of training, send a notification to the guardian stating "Training passed, travel permitted";

[0148] 14) After training ends, send a destination risk warning to the guardian.

[0149] In the above embodiments, dynamic VR modeling based on real-world environmental information differs from existing VR training based on preset virtual scenes. This application utilizes real-time or near-real-time environmental information (including images, videos, audio, etc.) of the target destination to dynamically construct a highly realistic VR training environment, and specifically introduces common dynamic stressors (such as sudden high-decibel noise, bright flashing lights) and typical social situations (such as shopping, queuing) that may cause discomfort in children with ASD within this environmental category. Multimodal fusion adaptive quantitative assessment and early warning: During VR training, multiple physiological signals (EEG, EOG, GSR) and behavioral information (voice, facial expressions, movements) of the child are simultaneously collected. By extracting and jointly analyzing these multimodal features, the system can objectively and quantitatively assess the child's adaptability, integration level, and potential discomfort level in the simulated environment. When the analysis results indicate a significant risk of discomfort, the system can issue preventative warnings before actual travel, guiding caregivers to take appropriate measures. The progressive training management model, based on the quantitative performance of multimodal data of patients, divides VR training into a multi-stage progressive mode. Each stage sets clear quantitative indicators (physiological stability, task completion, social cooperation). The "advancement / regression" is dynamically triggered according to the patient's continuous training performance, so as to achieve a precise match between training difficulty and the patient's adaptability.

[0150] This system achieves a closed loop through the collaborative work of six core modules: personalized threshold adaptation, hierarchical modeling of real-world scenarios, progressive training, data collection, quantitative assessment, and risk warning. It significantly improves the realism and adaptability of scenarios, with multiple VR scene versions generated based on real-world environmental data. Stressors are strongly correlated with scenarios, resolving the existing problem of "scenario disconnect" in training. The four-tiered scenario version design adapts to different training stages. It provides dual guarantees of training efficiency and safety, with progressive training and a dynamic switching mechanism to avoid resistance or ineffective training due to mismatched difficulty. A rollback mechanism reduces training risks and improves the child's adaptability to the target environment. The system ensures objectivity in assessment and warning, with multimodal data-driven quantitative assessment replacing subjective judgment and quantitatively controlling the switching of VR scene versions and training stages. It balances universality and personalization, supporting various outdoor scenarios such as supermarkets, hospitals, and parks. It can adjust environmental comfort thresholds and stable physiological ranges based on the child's historical behavioral information and clinical data to adapt to the personalized needs of different users.

[0151] In another embodiment, this application also provides a progressive training device pre-adapted to a real-world scenario, the device comprising:

[0152] The modeling module is used to acquire scene information of real scenes, split the scene information into elements at different levels, and perform progressive hierarchical modeling based on the elements at different levels after splitting to construct multi-level virtual scenes that simulate real environments at different levels.

[0153] The multimodal data synchronous acquisition module is used to acquire heterogeneous feature representations when a user wears a training device and performs pre-adaptation in the multi-level virtual scene;

[0154] The multimodal feature fusion analysis module is used to calculate the user's sensing parameters and / or interaction parameters based on the heterogeneous feature expression;

[0155] The training interaction module is used to dynamically switch the levels of a multi-level virtual scene based on the user's sensor parameters and / or interaction parameters, and to perform pre-adaptive training in the multi-level virtual scene after the switch.

[0156] See Figure 7 The device may further include a personalized threshold adaptation module for VR device adaptation correction, a modeling module (specifically a VR modeling module) for constructing multi-level progressive virtual scenes, and a training interaction module (specifically a VR interaction module) for training in different levels of virtual scenes and acquiring training data. A multimodal data synchronization acquisition module collects multimodal heterogeneous data from the user during training, and a multimodal feature fusion analysis module analyzes the data to assess the user's training status and adaptively switch training levels. Furthermore, the device can output evaluation results and provide risk warnings and feedback on stage-by-stage labeling progress via an early warning module.

[0157] For a detailed description of the progressive training device for pre-adapting to real-world scenarios provided by the present invention, please refer to the embodiment of the progressive training method for pre-adapting to real-world scenarios mentioned above, which will not be repeated here.

[0158] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a progressive training method for pre-adapting to a real-world scenario as described in any of the above embodiments.

[0159] This application also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of a progressive training method for pre-adapting to a real-world scenario as described in any of the above embodiments.

[0160] It is understood that the computer device provided in this application can be a server, and its internal structure diagram can be as follows: Figure 8As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The database stores relevant data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the method provided in this application.

[0161] Those skilled in the art will understand that Figure 8 The structures shown are merely block diagrams of a portion of the structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. The computer device may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program may include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0162] It should be understood that the processor mentioned in the embodiments of this application 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. A general-purpose processor can be a microprocessor or any conventional processor.

[0163] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0164] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A progressive training device pre-adapted to real-world scenarios, characterized in that, The device includes: The modeling module is used to acquire scene information of a real-world scenario, and to decompose the environmental elements and social context elements in the scene information according to their levels to obtain environmental elements and social context elements at different levels. Virtual scenes at corresponding levels are constructed using these environmental elements and social context elements. These different levels of virtual scenes are used to progressively simulate real-world scenarios, and as progressive training progresses, the virtual scenes continuously approximate the real-world scenarios. The real-world scenario is the user's destination scenario. The multimodal data synchronous acquisition module is used to acquire heterogeneous feature representations when users wear training devices and perform pre-adaptation in multi-level virtual scenes; The multimodal feature fusion analysis module is used to calculate the user's sensing parameters and / or interaction parameters based on the heterogeneous feature expression; The training interaction module is used to dynamically switch the levels of multi-level virtual scenes based on the user's sensory parameters and / or interaction parameters, and to perform pre-adaptation training in the multi-level virtual scenes after the switch. It includes a rollback mechanism; if at least one of the user's physiological stability characteristic interval, stress response degree, task completion degree, indicator stability degree, or care assessment degree corresponding to the sensory parameters and / or interaction parameters triggers a high-risk threshold in the corresponding level of the virtual environment, the pre-adaptation training to the real-world scene is terminated. The calculation steps for the physiological stability characteristic interval include: obtaining a first parameter of the user in the current scene based on safety environment standards, the first parameter including an environmental comfort reference value; obtaining the scene risk correction weight corresponding to the current scene category; adjusting the first parameter according to the scene risk correction strategy, the scene risk correction weight, and the sensitivity correction strategy to obtain the environmental critical condition; obtaining the user's physiological stability characteristics in the current scene category and current activity state; correcting the physiological stability characteristics using the scene risk correction strategy to obtain the user's physiological stability characteristic interval; collecting the user's electroencephalogram (EEG) signals; calculating the user's adaptation correction value to the training device based on the EEG signals; and correcting the physiological stability characteristic interval based on the adaptation correction value.

2. The apparatus according to claim 1, characterized in that, The device is also used for: Acquire heterogeneous feature representations of users during pre-adaptation in the current level of the virtual scene; Based on the sensing parameters and interaction parameters in the heterogeneous feature representation, the user's fitness level in the current level of the virtual scene is determined, and the level of the virtual environment in which the user is located is dynamically switched according to the fitness level.

3. The apparatus according to claim 2, characterized in that, The device is also used for: Determine the environmental adaptability corresponding to the user's physiological stability based on the user's sensor parameters and / or interaction parameters; Determine the stress response level corresponding to the user's behavioral performance based on the user's sensor parameters and / or interaction parameters; The task completion rate corresponding to the quality of the user's task completion is determined based on the user's sensor parameters and / or interaction parameters. Determine the stability of the index corresponding to the user stability verification based on the user's sensing parameters and / or interaction parameters; Determine the care assessment level corresponding to the collaboration between the user and the caregiver based on the user's sensor parameters and / or interaction parameters; The user's adaptability in the current level of the virtual scene is determined based on at least one of the following: environmental adaptability, stress response, task completion, indicator stability, and care assessment.

4. The apparatus according to claim 2, characterized in that, The device is also used for: If the user's fitness level in the current virtual scene, as determined by the sensing parameters and / or interaction parameters, meets the switching conditions, then the user is switched from the current virtual scene to the next level virtual scene. This allows the user to undergo virtual sensory stimulation training in the next level virtual scene to adapt to the real scene. The virtual scene level is iteratively adjusted based on the user's fitness level until the user completes virtual sensory stimulation training in the highest level virtual scene to adapt to the real scene or the training is terminated. If the user's adaptation level in the current virtual scene, as determined by the sensing parameters and / or interaction parameters, does not meet the switching conditions, the system continues to check whether the user has a rollback record that has reached the preset maximum number of rollbacks in the current virtual scene. If so, a risk warning is issued or training is terminated. If not, the user continues to undergo virtual sensory stimulation training in the current virtual scene to adapt to the real scene, and the level of the virtual scene is iteratively adjusted according to the user's adaptation level until the user completes virtual sensory stimulation training in the highest level virtual scene to adapt to the real scene or training is terminated.

5. The apparatus according to claim 1, characterized in that, The device is also used for: When a user is first placed in a virtual scene at the key level for pre-adaptation to a real scene for virtual sensory stimulation training, no sensory hypersensitivity items are set for the user in the environmental elements at the key level during the construction of the virtual scene at the key level. When a user is not in a virtual scene at the key level for the first time to conduct virtual sensory stimulation training to adapt to the real scene, the user's sensory hypersensitivity items are dynamically set in the environmental elements of the key level to construct the virtual scene at the key level, wherein the key level is the second-highest level.

6. The apparatus according to claim 1, characterized in that, The device is also used for: Collect resting-state EEG signals from users when they are not wearing training devices, and calculate the first set of characteristic parameters; Collect resting-state EEG signals of children after they wear the training device, and calculate the second set of characteristic parameters; The user's fitness correction value for the training device is determined based on the difference between the first set of feature parameters and the second set of feature parameters. A scaling factor is determined based on the fitness correction value, and the physiological stability feature interval is corrected based on the scaling factor.

Images