A child intelligent school uniform collar badge early warning method, system, terminal and storage medium based on AI feedback closed loop

Abstract

The present application relates to the technical field of intelligent wearable devices, and discloses a child intelligent school uniform epaulet early warning method and system based on AI feedback closed loop, a terminal and a storage medium. The method comprises the following steps: collecting children's eye behavior data and environmental data through the school uniform epaulet; pre-processing the data, and obtaining the current visual risk state of the children by the edge AI model of the epaulet; according to the state, controlling the feedback module on the epaulet to perform a warning action according to the current feedback strategy, and generating a warning log; obtaining the children's behavior feedback data after the warning, and uploading the warning log and the children's behavior feedback data to the cloud; the cloud evaluates the warning effect according to the children's behavior feedback data, and iteratively optimizes the feedback strategy by using a reinforcement learning algorithm, and after obtaining the target feedback strategy, the target feedback strategy is updated to the epaulet. The present application realizes all-weather, closed-loop and self-adaptive children's eye behavior management through the school uniform epaulet, and improves the effectiveness and compliance of intervention.

Description

Technical Field

[0001] This invention relates to the field of smart wearable device technology, and in particular to a method, system, terminal, and computer-readable storage medium for early warning of children's smart school uniform badges based on AI feedback closed loop. Background Technology

[0002] Currently, products on the market for monitoring children's eye behavior and protecting their vision mainly fall into the following categories: First, physical posture correctors, which rely on mechanical structures or simple tilt sensors for reminders. These have limited functionality, cannot recognize complex head postures, viewing distances, and changes in ambient light, and are uncomfortable to wear, leading to low child compliance. Second, smart desk lamps or learning machines with integrated cameras. While they can monitor posture and viewing distance, these devices are fixed and cannot move with the child, limiting the monitoring range. Furthermore, their warning methods are mostly fixed-frequency and fixed-intensity sound and light alarms, lacking personalized adjustment capabilities. Third, smart bracelets or watches. Some products have posture monitoring functions, but they struggle to accurately identify head and eye postures and cannot perceive actual viewing distances and ambient light intensity.

[0003] Existing technologies generally suffer from the following drawbacks: they have a single perception dimension, failing to simultaneously acquire multimodal data such as eye distance, head posture, torso posture, and ambient lighting; their recognition accuracy and real-time performance are insufficient, easily leading to false alarms or missed alarms in dynamic scenarios such as children's daily learning and movement; their intervention methods are rigid and non-personalized, unable to dynamically adjust the intensity and method of warnings based on children's individual differences and real-time reactions, and long-term use can easily lead to "reminder fatigue" or resistance; they lack an effective closed-loop optimization mechanism, making it difficult to quantify and evaluate the intervention effect, and the warning strategy cannot be continuously improved, resulting in low efficiency in behavior correction; in addition, existing devices are inconvenient to wear or have limited applicable scenarios, making it difficult to achieve all-weather, non-intrusive monitoring and management.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide a method, system, terminal, and computer-readable storage medium for early warning of children's smart school uniform badges based on AI feedback closed loop. This invention aims to solve the problems existing in children's eye behavior monitoring and early warning technologies, such as single perception modality, insufficient real-time performance and recognition accuracy, rigid intervention methods and inability to personalize and adaptively adjust, lack of closed-loop optimization mechanism based on behavioral feedback, poor device wearing compliance, and difficulty in achieving all-weather, non-intrusive monitoring.

[0006] To achieve the above objectives, this invention provides a method for early warning of children's smart school uniform badges based on AI feedback closed loop. The method includes the following steps: By integrating a multimodal sensing module into the school uniform badge, children's eye-use behavior data and environmental data are collected in real time; The child's eye-use behavior data and the environmental data are preprocessed, and reasoning is performed using the edge AI model deployed on the school uniform brooch to obtain the child's current visual risk status. Based on the visual risk status, the feedback module on the school uniform lapel is controlled to perform the corresponding warning action according to the currently effective feedback strategy, and a warning log is generated. The warning log records the visual risk status and the warning action. Obtain child behavior feedback data after the warning, and upload the warning log and the child behavior feedback data to the cloud platform; The cloud platform evaluates the early warning effect based on the child's behavioral feedback data, and uses a reinforcement learning algorithm to iteratively optimize the feedback strategy based on the early warning log and the early warning effect to obtain a target feedback strategy. The target feedback strategy is then sent to the school uniform badge to update the currently effective feedback strategy.

[0007] Furthermore, the multimodal sensing module integrated into the school uniform lapel pin collects children's eye-use behavior data and environmental data in real time, specifically including: The system uses a miniature camera integrated into a multimodal perception module on the school uniform lapel to capture a sequence of environmental images of the reading and writing area in front of the child. The system also uses an inertial sensor in the multimodal perception module to capture data on the child's torso movement posture. The environmental image sequence and the torso movement posture data are then used as data on the child's eye-use behavior. The ambient light sensor in the multimodal sensing module collects the light intensity value of the current read / write area and uses the light intensity value as environmental data.

[0008] Furthermore, the preprocessing of the children's eye-use behavior data and the environmental data specifically includes: The environmental image sequence in the children's eye use behavior data is denoised and enhanced, and the trunk motion posture data in the children's eye use behavior data is smoothed by Kalman filtering to obtain standardized behavior data; the standardized behavior data and the environmental data are time-stamp aligned and normalized to generate a multimodal input feature vector for model inference.

[0009] Furthermore, the process of inferring the child's current visual risk status through an AI model deployed at the edge of the school uniform badge specifically includes: The multimodal input feature vector is input into the edge AI model deployed on the school uniform lapel, and the edge AI model is used to perform monocular depth estimation on the environmental image sequence to obtain the eye distance estimation value; The edge AI model is used to detect and track read / write targets in the environmental image sequence to obtain the head pose deflection angle. The torso tilt angle is obtained by performing posture calculation on the torso motion posture data using the edge AI model. Based on the estimated eye distance, the head posture deflection angle, the torso tilt angle, and the light intensity value in the environmental data, the risk type and risk level are determined by a preset risk assessment rule, and the risk type and risk level are combined into the visual risk state.

[0010] Furthermore, based on the visual risk status, the step of controlling the feedback module on the school uniform lapel to execute corresponding early warning actions according to the currently effective feedback strategy, and generating an early warning log, specifically includes: Query the target feedback parameters that match the visual risk status in the currently effective feedback strategies, and drive the feedback module on the school uniform lapel to execute the corresponding intensity and type of warning action based on the target feedback parameters; The warning action, the visual risk status, and the target feedback parameters are recorded, and the recorded warning action, the visual risk status, and the target feedback parameters are encapsulated into a warning log.

[0011] Furthermore, the cloud platform evaluates the early warning effect based on the child's behavioral feedback data, specifically including: The cloud platform calculates the response time for the child's posture to return to normal after the warning, as well as the frequency of recurrence of undesirable behavior within a preset time period after the warning, based on the child's behavioral feedback data. The response time and the recurrence frequency of the adverse behavior are substituted into a preset effect evaluation function to calculate a quantitative reward value that characterizes the effectiveness of the intervention, and the quantitative reward value is used as the evaluation result of the early warning effect.

[0012] Furthermore, the step of iteratively optimizing the feedback strategy using a reinforcement learning algorithm based on the warning log and the warning effect to obtain a target feedback strategy, distributing the target feedback strategy to the school uniform badge, and updating the currently effective feedback strategy specifically includes: Using the visual risk state in the warning log as the state space, the warning action in the warning log as the action space, and the warning effect as the reward signal, a policy network is trained through a reinforcement learning algorithm to output the optimal feedback parameter configuration that maximizes the cumulative reward, and the optimal feedback parameter configuration is determined as the target feedback policy. The target feedback strategy is sent to the main control unit of the school uniform badge via a network transmission protocol, and the main control unit replaces the currently effective feedback strategy stored locally with the target feedback strategy.

[0013] Furthermore, to achieve the above objectives, the present invention also provides a children's intelligent school uniform badge early warning system based on AI feedback closed loop. This system is used to implement the children's intelligent school uniform badge early warning method based on AI feedback closed loop as described above. The system includes: The data acquisition module is used to collect children's eye-use behavior data and environmental data in real time through a multimodal sensing module integrated into the school uniform badge; The risk assessment module is used to preprocess the children's eye behavior data and the environmental data, and to infer the children's current visual risk status by using the edge AI model deployed on the school uniform badge. The early warning execution module is used to control the feedback module on the school uniform lapel to execute the corresponding early warning action according to the visual risk status and the currently effective feedback strategy, and generate an early warning log, wherein the content recorded in the early warning log includes the visual risk status and the early warning action; The data upload module is used to acquire child behavior feedback data after the warning and upload the warning log and the child behavior feedback data to the cloud platform. The strategy optimization module is used to evaluate the early warning effect based on the child's behavioral feedback data through the cloud platform, and to iteratively optimize the feedback strategy using a reinforcement learning algorithm based on the early warning log and the early warning effect to obtain a target feedback strategy. The target feedback strategy is then sent to the school uniform badge to update the currently effective feedback strategy.

[0014] Furthermore, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and an AI-based intelligent school uniform badge warning program stored in the memory and executable on the processor, wherein when the AI-based intelligent school uniform badge warning program is executed by the processor, it implements the steps of the AI-based intelligent school uniform badge warning method described above.

[0015] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a children's smart school uniform badge warning program based on AI feedback closed loop, and when the children's smart school uniform badge warning program based on AI feedback closed loop is executed by a processor, it implements the steps of the children's smart school uniform badge warning method based on AI feedback closed loop as described above.

[0016] The beneficial effects of this invention are as follows: By integrating a multimodal perception module into the school uniform lapel pin, this invention enables real-time, all-weather, and seamless collection of children's eye-use behavior and environmental data; preprocessing the collected data and using an edge AI model deployed on the lapel pin for inference achieves low-latency, high-precision real-time visual risk assessment; controlling the feedback module to execute corresponding warning actions and generate warning logs based on the currently effective feedback strategy enables immediate, personalized, and traceable intervention responses; acquiring children's behavioral feedback data after warnings and uploading it along with the warning logs to the cloud platform achieves cloud-based aggregation of multi-source behavioral data; the cloud platform evaluates the warning effect based on the feedback data and iteratively optimizes the feedback strategy using reinforcement learning algorithms, then distributes the optimized strategy back to the lapel pin to update the effective strategy, thereby achieving an adaptive, continuously evolving, personalized closed-loop optimization effect, significantly improving the long-term effectiveness of behavioral interventions and user compliance. Attached Figure Description

[0017] Figure 1 This is a flowchart of a preferred embodiment of the children's smart school uniform badge early warning method based on AI feedback closed loop of the present invention; Figure 2 This is a structural diagram of a preferred embodiment of the children's intelligent school uniform badge early warning system based on AI feedback closed loop of the present invention; Figure 3 This is a structural diagram of a preferred embodiment of the terminal of the present invention. Detailed Implementation

[0018] This application provides a method, system, terminal, and storage medium for early warning of children's smart school uniform badges based on AI feedback closed loop. To make the purpose, technical solution, and effects of this application clearer and more explicit, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.

[0019] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0020] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0021] The preferred embodiment of the present invention describes a children's smart school uniform badge early warning method based on AI feedback closed loop, such as... Figure 1 As shown, the AI-based intelligent school uniform badge early warning method for children includes the following steps: S10. Through the multimodal sensing module integrated into the school uniform lapel, children's eye behavior data and environmental data are collected in real time.

[0022] The purpose of this step is to use miniaturized, low-power sensor arrays embedded in school uniform badges to continuously and stably capture children's eye-use behavior characteristics and ambient lighting conditions without interfering with their normal activities, providing raw data input for edge AI risk assessment, thereby achieving all-weather, non-intrusive data collection.

[0023] In this embodiment, step S10 specifically includes: S11. Using a miniature camera in a multimodal perception module integrated on the school uniform lapel, a sequence of environmental images of the reading and writing area in front of the child is collected. Inertial sensors in the multimodal perception module are used to collect data on the child's torso movement posture. The environmental image sequence and the torso movement posture data are used as the child's eye behavior data.

[0024] Specifically, the ultra-miniature low-power camera uses a global shutter CMOS sensor, such as the OmniVision OV7251, with a resolution of no less than 640x480 and a frame rate of 30fps. It integrates an image signal processor (ISP) and features high dynamic range (HDR) and low-light performance. The lens is wide-angle (e.g., a field of view (FOV) of 90-120 degrees), and the camera lens is oriented towards the child's front to ensure stable capture of environmental images of the reading and writing area, such as the book the child is reading or the desktop they are looking at. The camera module size is controlled within 5x5x5mm. The inertial measurement unit (IMU) is a 6-axis IMU, such as the STMicroelectronics LSM6DSOX, which integrates a three-axis accelerometer and a three-axis gyroscope for real-time monitoring of the tilt angle, motion state, and vibration frequency of the child's torso. The system uses a timestamp synchronization mechanism to align the environmental image sequence captured by the camera with the torso posture data captured by the IMU, together forming the child's eye behavior data.

[0025] S12. The ambient light sensor in the multimodal sensing module collects the light intensity value of the current read / write area and uses the light intensity value as environmental data.

[0026] Specifically, the ambient light sensor is a high-precision sensor, such as the AMS AS7341 or Vishay VEML6030, capable of accurately measuring the ambient light intensity (in lux) and color temperature. This sensor is mounted on the front or side of the lapel pin to avoid obstruction by clothing, ensuring that the collected light data accurately represents the actual ambient brightness of the child's reading and writing area. Sensor data is reported to the main control unit in real time via an Inter-Integrated Circuit (I2C) interface.

[0027] It should be noted that, to ensure a balance between continuous data acquisition and energy efficiency, all the aforementioned sensors support low-power sleep or wake-up mechanisms. When the IMU detects that the child is stationary or in a non-learning state (such as lying down or walking) for more than a preset threshold (e.g., 5 minutes), the main control unit will control the camera and ambient light sensor to enter sleep mode to extend battery life. When the IMU detects a change in posture again or a periodic wake-up timer is triggered, the system automatically resumes full-speed data acquisition.

[0028] S20. The child's eye-use behavior data and the environmental data are preprocessed, and the child's current visual risk status is obtained by reasoning through the edge AI model deployed on the school uniform badge.

[0029] The purpose of this step is to clean, enhance, and extract features from the raw multimodal data at the edge computing device, and to use a lightweight deep learning model for real-time inference to transform the raw sensor data into a quantifiable visual risk level. This reduces the data transmission bandwidth requirements and cloud computing pressure while ensuring user privacy and real-time response.

[0030] In step S20, the preprocessing of the child's eye-use behavior data and the environmental data specifically includes: S211. The environmental image sequence in the children's eye use behavior data is subjected to denoising and enhancement processing, and the trunk movement posture data in the children's eye use behavior data is subjected to Kalman filtering smoothing processing to obtain standardized behavior data.

[0031] Specifically, after receiving the raw image, the main control unit (e.g., the ESP32-S3 microcontroller) first calls the built-in image preprocessing library to perform Gaussian filtering for noise reduction and histogram equalization for enhancement, in order to reduce the effects of low light or motion blur. At the same time, the raw three-axis acceleration and angular velocity data output by the IMU are fused and smoothed using the Kalman filter algorithm to eliminate high-frequency noise and instantaneous jitter, resulting in stable body attitude angles (pitch angle, roll angle, yaw angle).

[0032] S212. The standardized behavioral data and the environmental data are time-stamp aligned and normalized to generate a multimodal input feature vector for model inference.

[0033] Specifically, the system precisely aligns the preprocessed image frames, torso pose angle values, and ambient light intensity values ​​according to their respective timestamps. Subsequently, the environmental image data is scaled to the model input size (e.g., 224x224 pixels), and the pose angle and light intensity are normalized (e.g., scaled to the [0,1] interval), concatenated into a multimodal input feature vector, and fed into the edge AI model.

[0034] In step S20, the step of obtaining the child's current visual risk status by reasoning through the edge AI model deployed on the school uniform brooch specifically includes: S221. Input the multimodal input feature vector into the edge AI model deployed on the school uniform patch, perform monocular depth estimation on the environmental image sequence through the edge AI model to obtain the eye distance estimation value; detect and track reading and writing targets in the environmental image sequence through the edge AI model to obtain the head posture deflection angle; perform posture calculation on the torso motion posture data through the edge AI model to obtain the torso tilt angle.

[0035] Specifically, the edge AI model is a lightweight deep learning model deployed on a microcontroller unit (MCU), comprising a monocular depth estimation subnetwork and an object detection and tracking subnetwork. This embodiment uses a lightweight model based on the MobileNetV3 architecture and has been converted to TFLite or ONNX format. The model size is controlled to within a few MB. After loading the model, forward inference is performed on the input multimodal feature vectors, outputting the following key metrics: Eye distance estimation: The environmental image sequence is processed by the monocular depth estimation sub-network in the edge AI model to estimate the depth information of each pixel in the environmental image from the camera. The average depth value of the reading and writing target area is extracted from the estimated depth information. Based on the average depth value and the camera calibration parameters, the distance from the child's eyes to the reading and writing target (unit: centimeters) is inferred.

[0036] Head posture yaw angle: The target detection and tracking sub-network in the edge AI model is used to detect and track the reading and writing targets in the environmental image sequence. The position changes, size changes and perspective distortion of the reading and writing targets in the environmental image sequence are analyzed. Based on the position changes, size changes and perspective distortion, combined with the camera calibration parameters, the pitch angle (head down or head up), yaw angle (left or right tilt) and roll angle of the child's head relative to the reading and writing target are inferred.

[0037] Trunk tilt angle: directly taken from the preprocessed IMU posture calculation results, reflecting the degree of curvature or lateral deviation of the child's spine. S222, based on the estimated eye distance, the head posture deflection angle, the trunk tilt angle, and the light intensity value in the environmental data, the risk type and risk level are determined by a preset risk assessment rule, and the risk type and risk level are combined into the visual risk state.

[0038] Specifically, the rule engine or a small neural network classifier inside the main control unit compares the above output value with multiple preset thresholds. For example: If the viewing distance is less than 25 centimeters, the risk type is determined to be "too close to the eyes".

[0039] If the head pitch angle is greater than 20 degrees or the torso roll angle is greater than 15 degrees, the risk type is determined to be "incorrect posture" (which can be further subdivided into "head down", "tilting to the side", etc.).

[0040] If the ambient light intensity is less than 300 Lux or greater than 1500 Lux, the risk type is determined to be "unsuitable lighting environment".

[0041] Each risk type is further divided into three risk levels: "mild," "moderate," and "severe." For example, a viewing distance of 20-25cm is considered mild, 15-20cm is moderate, and less than 15cm is severe. Finally, all triggered risk types and their levels are combined into a "visual risk status" vector for the current moment.

[0042] It should be noted that the edge AI model is not static. The cloud platform periodically trains a better base model using massive amounts of anonymized user data through federated learning (FL) or transfer learning (TL) techniques. While protecting user privacy, it performs personalized fine-tuning for each child's specific behavioral habits and then silently updates the model to the device via Wi-Fi, thereby continuously improving the model's recognition accuracy.

[0043] S30. Based on the visual risk status, control the feedback module on the school uniform lapel to perform the corresponding warning action according to the currently effective feedback strategy, and generate a warning log, wherein the content recorded in the warning log includes the visual risk status and the warning action.

[0044] The purpose of this step is to translate the visual risks assessed by AI into gentle and effective physical feedback (light, vibration, sound) in real time, enabling immediate behavioral intervention, while recording the complete context of this warning.

[0045] In this embodiment, step S30 specifically includes: S31. Query the target feedback parameters that match the visual risk status in the currently effective feedback strategy, and drive the feedback module on the school uniform lapel to perform a warning action of corresponding intensity and type according to the target feedback parameters.

[0046] Specifically, the main control unit maintains a feedback policy table (or policy network) in its memory, which defines the mapping relationship between different risk types and levels and feedback actions. For example: Mild risk: Drives a miniature breathing light (RGB LED) to display yellow and blinks in a "breathing" manner at a low frequency (1Hz). Simultaneously, drives a miniature linear resonant actuator (LRA) to emit a slight vibration lasting 0.5 seconds.

[0047] Moderate risk: The breathing light displays orange and flashes at a faster frequency of 2Hz. The vibration motor vibrates continuously at a moderate intensity for 1 second. If it continues for more than 10 seconds, a low-volume prompt tone such as "Please sit up straight" or "Keep your eyes away" will be played through a miniature speaker.

[0048] Severe risk: The breathing light displays red and flashes rapidly (4Hz). The vibration motor vibrates intermittently at high intensity (vibrate for 0.5 seconds, stop for 0.5 seconds). Simultaneously, a real-time alert notification is sent to the paired parent or teacher's app via Bluetooth Low Energy (BLE).

[0049] The miniature breathing light uses PWM dimming technology to achieve a smooth color transition. The miniature vibration motor uses an LRA (Low-Range Resonance Motor) because of its fast response speed, clear vibration direction, and low power consumption.

[0050] S32. Record the warning action, the visual risk status, and the target feedback parameters, and encapsulate the recorded warning action, the visual risk status, and the target feedback parameters into a warning log.

[0051] Specifically, after each warning action is executed, the system encapsulates the current timestamp, the visual risk status (including risk type and level), the target feedback parameters (such as LED color, brightness, flashing mode, vibration intensity, and frequency), and the action type into a structured warning log, and temporarily stores it in the local Flash memory or Micro SD card.

[0052] S40. Obtain the child's behavior feedback data after the warning, and upload the warning log and the child's behavior feedback data to the cloud platform.

[0053] The purpose of this step is to upload the behavioral intervention records and children's actual response data accumulated on the platform to the cloud, providing necessary state-action-reward samples for big data analysis and reinforcement learning model training in the cloud, thereby driving the continuous evolution of the entire feedback loop.

[0054] In this embodiment, when the wireless network (Wi-Fi) is available, the main control unit uploads the locally cached warning logs and the child's behavioral feedback data collected after the warning (including the sequence of posture changes, the sequence of changes in eye distance, and the time required for risk clearance within 10 seconds after the warning) to the cloud platform via Hypertext Transfer Protocol Secure (HTTPS) or Message Queuing Telemetry Transport (MQTT). If the network is unavailable, the data will continue to be cached locally and will be resumed after the network is restored.

[0055] It should be noted that, to protect children's privacy, feature extraction is completed on the uploaded image data at the badge end (only structured data such as pose angle and distance are uploaded). Original image frames are not uploaded by default, except for a small number of anonymized samples used for model optimization. Users can authorize whether to allow the uploading of anonymized low-resolution images for improving AI models via the app.

[0056] S50. The cloud platform evaluates the warning effect based on the child's behavior feedback data, and uses a reinforcement learning algorithm to iteratively optimize the feedback strategy based on the warning log and the warning effect to obtain a target feedback strategy. The target feedback strategy is then sent to the school uniform badge to update the currently effective feedback strategy.

[0057] The purpose of this step is to leverage the powerful computing capabilities and massive amounts of data from cloud computing to automatically learn the most effective personalized early warning strategies for each child through reinforcement learning algorithms, and to update the optimized strategies in a closed loop to edge devices, enabling the system to adaptively improve the long-term effectiveness of behavior correction.

[0058] In step S50, the cloud platform evaluates the early warning effect based on the child's behavioral feedback data, specifically including: S511. The cloud platform calculates the response time for the child's posture to return to normal after the warning, and the frequency of recurrence of bad behavior within a preset time period after the warning, based on the child's behavior feedback data.

[0059] Specifically, after receiving the data, the cloud platform first parses and stores it. For each warning log, two key indicators are extracted from the associated child behavior feedback data: Response Time (RT): The time (in seconds) elapsed from the start of the warning action until the child's posture or eye distance first returns to the normal threshold (e.g., eye distance greater than 30cm). The shorter the RT, the more timely and effective the warning intervention.

[0060] Relapse Frequency (RF): The number of times the same undesirable posture or behavior recurs within the next 5 minutes after the warning ends. The lower the RF, the more lasting the short-term corrective effect of the warning.

[0061] S512. Substitute the response duration and the recurrence frequency of the adverse behavior into a preset effect evaluation function to calculate a quantitative reward value that characterizes the effectiveness of the intervention, and use the quantitative reward value as the evaluation result of the early warning effect.

[0062] Specifically, the cloud platform designs a reward function, for example: Reward = α * (T)max - RT) -β* RF; Among them, T max This is the preset maximum tolerable response time (e.g., 10 seconds), and α and β are weighting coefficients. This function maps the child's "physiological-behavioral" response to a scalar reward value. The higher the reward value, the better the effect of this warning strategy (i.e., the specific combination of feedback parameters) on the current behavior correction of the child.

[0063] In step S50, the step of iteratively optimizing the feedback strategy using a reinforcement learning algorithm based on the warning log and the warning effect to obtain a target feedback strategy, distributing the target feedback strategy to the school uniform badge, and updating the currently effective feedback strategy specifically includes: S521. Using the visual risk state in the warning log as the state space, the warning action in the warning log as the action space, and the warning effect as the reward signal, a policy network is trained using a reinforcement learning algorithm to output the optimal feedback parameter configuration that maximizes the cumulative reward, and the optimal feedback parameter configuration is determined as the target feedback policy.

[0064] Specifically, the cloud platform constructs a reinforcement learning (RL) environment, in which: State Space: This includes visual risk type and level, child's historical behavioral characteristics (such as the rate of maintaining correct posture in the past hour), child's age, etc.

[0065] Action Space: This includes combinations of all adjustable parameters such as the color, brightness, and flashing mode of the breathing light; the intensity, frequency, and duration of the vibration motor; whether to play voice messages and the content of those messages; and whether to push notifications to the app.

[0066] Reward Signal: The quantified reward value calculated in step S512.

[0067] The cloud-based system employs Reinforcement Learning (RL) algorithms such as Deep Q-Network (DQN) or Proximal Policy Optimization (PPO) to train a policy network using massive amounts of tuple data (state, action, reward) collected from tens of thousands of child users. This policy network can output the optimal feedback parameter configuration that maximizes long-term cumulative reward (i.e., behavior correction effect) for a specific child and a specific visual risk state. This configuration constitutes the personalized target feedback policy for that child.

[0068] S522. The target feedback strategy is sent to the main control unit of the school uniform badge via a network transmission protocol, and the main control unit replaces the currently effective feedback strategy stored locally with the target feedback strategy.

[0069] Specifically, the cloud platform pushes the generated target feedback policy (in JSON format or binary protocol packet) to the corresponding school uniform badge via Wi-Fi. After receiving and verifying the integrity of the policy packet, the main control unit (MCU) on the badge end writes it to the policy storage area in local non-volatile memory (such as Flash) and atomically sets the pointer to the new policy, making it effective immediately. The next time the system makes an early warning decision, it will use the updated feedback policy.

[0070] Through the continuous cycle of S10 to S50 described above, this invention constructs a biofeedback closed loop of "perception-decision-execution-optimization". The system can not only monitor and intervene in children's eye-use behavior in real time, but also continuously evolve itself through reinforcement learning by interacting with children, providing each child with increasingly accurate, effective and personalized vision health management solutions, achieving a fundamental leap from passive reminders to proactive, adaptive behavior reshaping.

[0071] In summary, the AI-based feedback closed-loop early warning method for children's smart school uniform badges provided by this invention deeply integrates ultra-miniature multimodal sensors, edge AI inference, personalized non-invasive feedback, and cloud-based reinforcement learning optimization onto the school uniform badge as a carrier. This achieves all-weather, high-precision, closed-loop adaptive management of children's eye-use behavior and visual risk early warning, significantly improving the effectiveness and compliance of intervention.

[0072] Furthermore, such as Figure 2 As shown, based on the above-mentioned AI-based feedback closed-loop warning method for children's smart school uniform badges, this invention also provides a corresponding AI-based feedback closed-loop warning system for children's smart school uniform badges, the AI-based feedback closed-loop warning system for children's smart school uniform badges comprising: Data acquisition module 51 is used to collect children's eye use behavior data and environmental data in real time through a multimodal sensing module integrated on the school uniform badge; Risk calculation module 52 is used to preprocess the child's eye behavior data and the environmental data, and to infer the child's current visual risk status by using the edge AI model deployed on the school uniform badge. The early warning execution module 53 is used to control the feedback module on the school uniform lapel to execute the corresponding early warning action according to the visual risk status and the currently effective feedback strategy, and generate an early warning log, wherein the content recorded in the early warning log includes the visual risk status and the early warning action; Data upload module 54 is used to acquire child behavior feedback data after the warning, and upload the warning log and the child behavior feedback data to the cloud platform; The strategy optimization module 55 is used to evaluate the early warning effect based on the child's behavior feedback data through the cloud platform, and to iteratively optimize the feedback strategy using a reinforcement learning algorithm based on the early warning log and the early warning effect to obtain a target feedback strategy. The target feedback strategy is then sent to the school uniform badge to update the currently effective feedback strategy.

[0073] Furthermore, such as Figure 3 As shown, based on the above-mentioned AI-based intelligent school uniform badge early warning method and system, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 3 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0074] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard drive or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard drive, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as program code installed on the terminal. The memory 20 may also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a children's smart school uniform badge warning program 40 based on AI feedback closed loop. This children's smart school uniform badge warning program 40 based on AI feedback closed loop can be executed by the processor 10, thereby implementing the children's smart school uniform badge warning method based on AI feedback closed loop in this application.

[0075] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the AI-based feedback closed-loop children's smart school uniform badge early warning method.

[0076] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components of the terminal communicate with each other via a system bus.

[0077] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a children's smart school uniform badge warning program based on AI feedback closed loop, and the children's smart school uniform badge warning program based on AI feedback closed loop, when executed by a processor, implements the steps of the children's smart school uniform badge warning method based on AI feedback closed loop as described above.

[0078] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0079] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0080] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for early warning of children's smart school uniform badges based on AI feedback closed loop, characterized in that, The AI-based intelligent school uniform badge early warning method for children includes the following steps: By integrating a multimodal sensing module into the school uniform badge, children's eye-use behavior data and environmental data are collected in real time; The child's eye-use behavior data and the environmental data are preprocessed, and reasoning is performed using the edge AI model deployed on the school uniform brooch to obtain the child's current visual risk status. Based on the visual risk status, the feedback module on the school uniform lapel is controlled to perform the corresponding warning action according to the currently effective feedback strategy, and a warning log is generated. The warning log records the visual risk status and the warning action. Obtain child behavior feedback data after the warning, and upload the warning log and the child behavior feedback data to the cloud platform; The cloud platform evaluates the early warning effect based on the child's behavioral feedback data, and uses a reinforcement learning algorithm to iteratively optimize the feedback strategy based on the early warning log and the early warning effect to obtain a target feedback strategy. The target feedback strategy is then sent to the school uniform badge to update the currently effective feedback strategy.

2. The method for early warning of children's smart school uniform badges based on AI feedback closed loop according to claim 1, characterized in that, The system utilizes a multimodal sensing module integrated into the school uniform lapel to collect real-time data on children's eye-use behavior and environmental data, specifically including: The system uses a miniature camera integrated into a multimodal perception module on the school uniform lapel to capture a sequence of environmental images of the reading and writing area in front of the child. The system also uses an inertial sensor in the multimodal perception module to capture data on the child's torso movement posture. The environmental image sequence and the torso movement posture data are then used as data on the child's eye-use behavior. The ambient light sensor in the multimodal sensing module collects the light intensity value of the current read / write area and uses the light intensity value as environmental data.

3. The method for early warning of children's intelligent school uniform badges based on AI feedback closed loop as described in claim 1, characterized in that, The preprocessing of the children's eye-use behavior data and the environmental data specifically includes: The environmental image sequence in the children's eye use behavior data is denoised and enhanced, and the trunk movement posture data in the children's eye use behavior data is smoothed by Kalman filtering to obtain standardized behavior data. The standardized behavioral data and the environmental data are time-stamped and normalized to generate a multimodal input feature vector for model inference.

4. The method for early warning of children's intelligent school uniform badges based on AI feedback closed loop according to claim 3, characterized in that, The process of reasoning through an AI model deployed at the edge of the school uniform brooch to determine the child's current visual risk status specifically includes: The multimodal input feature vector is input into the edge AI model deployed on the school uniform lapel, and the edge AI model is used to perform monocular depth estimation on the environmental image sequence to obtain the eye distance estimation value; The edge AI model is used to detect and track read / write targets in the environmental image sequence to obtain the head pose deflection angle. The torso tilt angle is obtained by performing posture calculation on the torso motion posture data using the edge AI model. Based on the estimated eye distance, the head posture deflection angle, the torso tilt angle, and the light intensity value in the environmental data, the risk type and risk level are determined by a preset risk assessment rule, and the risk type and risk level are combined into the visual risk state.

5. The method for early warning of children's intelligent school uniform badges based on AI feedback closed loop according to claim 1, characterized in that, Based on the visual risk status, the system controls the feedback module on the school uniform lapel to execute corresponding early warning actions according to the currently effective feedback strategy, and generates an early warning log, specifically including: Query the target feedback parameters that match the visual risk status in the currently effective feedback strategies, and drive the feedback module on the school uniform lapel to execute the corresponding intensity and type of warning action based on the target feedback parameters; The warning action, the visual risk status, and the target feedback parameters are recorded, and the recorded warning action, the visual risk status, and the target feedback parameters are encapsulated into a warning log.

6. The method for early warning of children's intelligent school uniform badges based on AI feedback closed loop according to claim 1, characterized in that, The cloud platform evaluates the effectiveness of the early warning based on the child's behavioral feedback data, specifically including: The cloud platform calculates the response time for the child's posture to return to normal after the warning, as well as the frequency of recurrence of undesirable behavior within a preset time period after the warning, based on the child's behavioral feedback data. The response time and the recurrence frequency of the adverse behavior are substituted into a preset effect evaluation function to calculate a quantitative reward value that characterizes the effectiveness of the intervention, and the quantitative reward value is used as the evaluation result of the early warning effect.

7. The method for early warning of children's intelligent school uniform badges based on AI feedback closed loop according to claim 1, characterized in that, The step of iteratively optimizing the feedback strategy using a reinforcement learning algorithm based on the warning log and the warning effect to obtain a target feedback strategy, distributing the target feedback strategy to the school uniform badge, and updating the currently effective feedback strategy specifically includes: Using the visual risk state in the warning log as the state space, the warning action in the warning log as the action space, and the warning effect as the reward signal, a policy network is trained through a reinforcement learning algorithm to output the optimal feedback parameter configuration that maximizes the cumulative reward, and the optimal feedback parameter configuration is determined as the target feedback policy. The target feedback strategy is sent to the main control unit of the school uniform badge via a network transmission protocol, and the main control unit replaces the currently effective feedback strategy stored locally with the target feedback strategy.

8. A children's intelligent school uniform badge early warning system based on AI feedback closed loop, characterized in that, The AI-based intelligent school uniform badge early warning system is used to implement the AI-based intelligent school uniform badge early warning method as described in any one of claims 1-7, wherein the AI-based intelligent school uniform badge early warning system includes: The data acquisition module is used to collect children's eye-use behavior data and environmental data in real time through a multimodal sensing module integrated into the school uniform badge; The risk assessment module is used to preprocess the children's eye behavior data and the environmental data, and to infer the children's current visual risk status by using the edge AI model deployed on the school uniform badge. The early warning execution module is used to control the feedback module on the school uniform lapel to execute the corresponding early warning action according to the visual risk status and the currently effective feedback strategy, and generate an early warning log, wherein the content recorded in the early warning log includes the visual risk status and the early warning action; The data upload module is used to acquire child behavior feedback data after the warning and upload the warning log and the child behavior feedback data to the cloud platform. The strategy optimization module is used to evaluate the early warning effect based on the child's behavioral feedback data through the cloud platform, and to iteratively optimize the feedback strategy using a reinforcement learning algorithm based on the early warning log and the early warning effect to obtain a target feedback strategy. The target feedback strategy is then sent to the school uniform badge to update the currently effective feedback strategy.

9. A terminal, characterized in that, The terminal includes: a memory, a processor, and an AI-based intelligent school uniform badge warning program stored in the memory and executable on the processor. When the AI-based intelligent school uniform badge warning program is executed by the processor, it implements the steps of the AI-based intelligent school uniform badge warning method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a children's smart school uniform badge warning program based on AI feedback closed loop. When the children's smart school uniform badge warning program based on AI feedback closed loop is executed by a processor, it implements the steps of the children's smart school uniform badge warning method based on AI feedback closed loop as described in any one of claims 1-7.

Images