Pre-school fusion education behavior correction system and method based on AI empowerment

Abstract

The present application belongs to the technical field of behavior correction and educational informatization, and specifically relates to a pre-school fusion education behavior correction system and method based on AI empowerment, which comprises a data acquisition module, an AI analysis and processing module, a correction content library, an interactive execution module, a physiological monitoring module, an effect evaluation module and a teaching feedback module, and each module cooperates to realize a complete behavior correction process.The present application realizes accurate identification and analysis of problem behaviors of pre-school children, formulates individualized correction schemes in combination with individual characteristics of children, introduces AI technology and physiological monitoring means, improves the scientificity and objectivity of the correction process, solves the subjective defects of artificial evaluation, builds an integrated platform of teaching and practice, provides real cases and data support for behavior correction courses, improves the teaching effect, realizes dynamic tracking and quantitative evaluation of correction effect, timely adjusts the correction strategy, and improves the correction efficiency.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of behavior modification technology and educational informatization, specifically involving an AI-enabled behavior modification system and method for preschool inclusive education. It is mainly aimed at preschool inclusive education scenarios, combining AI technology to achieve personalized behavior modification, and is applicable to teaching practice in preschool education and inclusive education majors, as well as the work of correcting problem behaviors of special children and normal children. Background Technology

[0002] Currently, behavior modification technology is increasingly widely used in the field of inclusive preschool education, but existing technologies have many shortcomings. Traditional behavior modification methods rely heavily on manual observation and experience-based judgment to formulate plans, which suffers from strong subjectivity and insufficient personalization, making it difficult to accurately match the behavioral characteristics and correction needs of different children.

[0003] Existing correctional systems, such as some psychological correctional systems for special individuals, while incorporating mechanisms like physiological monitoring and question bank testing, are primarily applicable to adult special groups and do not fully consider the cognitive level and behavioral characteristics of preschool children, lacking adaptability to preschool inclusive education scenarios. Personalized correctional recommendation systems for incarcerated individuals, although possessing knowledge bases and recommendation modules, focus on crime causation analysis and resource recommendations, failing to meet the needs of preschool children's behavioral correction for real-time interaction, scenario simulation, and dynamic assessment.

[0004] Meanwhile, current behavioral modification teaching and practice in the field of preschool education suffers from problems such as rigid teaching content, case studies that are inconsistent with national and student conditions, and monotonous teaching methods. Furthermore, the modification process lacks scientific physiological data support and intelligent effectiveness evaluation methods, resulting in low efficiency, difficulty in quantifying effects, and an inability to provide accurate practical data feedback for teaching. These problems severely restrict the effectiveness of behavioral modification technology in the field of inclusive preschool education, necessitating a highly adaptable, intelligent, precise, and efficient behavioral modification system and method. Summary of the Invention

[0005] To address the shortcomings of existing behavior correction systems, such as insufficient adaptability to preschool inclusive education scenarios, low personalization, lack of intelligent assessment and feedback mechanisms, and disconnect between teaching and practice, this invention provides an AI-enabled preschool inclusive education behavior correction system and method. The aim is to: accurately identify and analyze problem behaviors in preschool children, and develop personalized correction plans based on individual child characteristics; introduce AI technology and physiological monitoring methods to improve the scientific rigor and objectivity of the correction process, overcoming the subjectivity of manual assessment; build an integrated teaching and practice platform to provide real-world cases and data support for behavior correction courses, thereby enhancing teaching effectiveness; and achieve dynamic tracking and quantitative evaluation of correction effects, allowing for timely adjustments to correction strategies and improving correction efficiency.

[0006] To achieve the above objectives, the present invention provides the following solution: The AI-enabled preschool inclusive education behavior correction system includes: a data acquisition module, an AI analysis and processing module, a correction content library, an interactive execution module, a physiological monitoring module, an effect evaluation module, and a teaching feedback module. The data acquisition module is used to collect basic information, behavioral data, and teaching scenario data of children to be corrected; The correction content library is used to store behavioral correction resources related to the field of preschool inclusive education; The AI ​​analysis and processing module is used to identify and extract features from the data collected by the data acquisition module, and generate a correction plan by combining it with behavior correction resources. The interactive execution module is used to enable teachers and parents to guide children to be corrected to participate in corrective training based on the corrective plan; The physiological monitoring module is used to collect physiological data of the child to be corrected during the correction process and transmit it to the effect evaluation module; The effect evaluation module is used to compare and quantify behavioral data, physiological data and baseline data before correction to obtain a correction effect score, and adjust the correction plan based on the correction effect score. The teaching feedback module is used to synchronize data, cases, and assessment results from the correction process to the behavior correction course teaching platform, providing practical case support and data resources for the course.

[0007] Preferably, the AI ​​analysis and processing module includes: a behavior recognition unit, a feature extraction unit, and a solution generation unit; The behavior recognition unit is used to automatically identify and classify behavioral data using computer vision technology and machine learning algorithms, and accurately determine the type of problematic behavior. The feature extraction unit is used to extract individual feature vectors of children from basic information and behavioral data, including behavioral preferences, emotional response patterns, and learning acceptance abilities; The solution generation unit is used to generate personalized correction solutions based on the child's individual feature vectors, combined with expert cases and resources in the correction content library, through collaborative filtering algorithms and decision tree models.

[0008] Preferably, the interactive execution module includes: a teacher terminal, a parent terminal, and a child interactive terminal; The teacher terminal is used to view children's data and correction plans, issue correction tasks, record the correction implementation process, and adjust correction strategies. Parent terminal is used to receive correction tasks from the home terminal and provide feedback on the child's behavior at home; The children's interactive terminal uses a fun interface design to present corrective content, guide children to participate in corrective training, and collect children's interactive feedback data in real time.

[0009] Preferably, the effect evaluation module includes: a data comparison unit, a quantitative scoring unit, and a strategy adjustment unit; The data comparison unit is used to compare behavioral and physiological data during the correction process with baseline data before correction. The quantitative scoring unit uses a percentage-based scoring method to calculate the correction effect score from three dimensions: degree of behavioral improvement, physiological stability, and interaction participation, based on the comparison results. The strategy adjustment unit automatically triggers the AI ​​analysis and processing module to regenerate or optimize the correction plan if the score is greater than or equal to a preset threshold, based on the correction effect score.

[0010] This invention also provides an AI-enabled method for behavioral correction in inclusive preschool education, which is implemented using the aforementioned system and includes: Step 1: Collect basic information, problem behavior data, and teaching scenario data of the children to be corrected through the data acquisition module. At the same time, collect baseline physiological data of the children to be corrected before correction through the physiological monitoring module. Step 2: Use the AI ​​analysis and processing module to automatically identify and classify the behavioral data and extract the individual feature vectors of children. Combine this with the behavioral correction resources in the correction content library to generate a correction plan. Step 3: Through the interactive execution module, teachers and parents guide children to participate in corrective training according to the correction plan, and collect interactive feedback data and physiological data simultaneously. Step 4: Compare behavioral and physiological data before and after correction through the effect evaluation module, calculate the correction effect score, and maintain or optimize the correction plan based on the correction effect score; Step 5: Synchronize case data and assessment reports from the correction process to the behavior correction course teaching platform through the teaching feedback module, update the teaching material library, and provide practical support for course teaching; Step 6: Repeat steps 3-5 until the child's correction effect score is ≤25 points for two consecutive times. Once this is completed, the final correction report will be generated and archived in the correction content library and teaching platform.

[0011] Preferably, step 2, which uses an AI analysis and processing module to automatically identify and classify behavioral data and extract individual child feature vectors, and combines this with behavioral correction resources in the correction content library to generate a correction plan, includes the following methods: The behavior recognition unit of the AI ​​analysis and processing module automatically identifies and classifies behavioral data to determine the type of problematic behavior. The feature extraction unit of the AI ​​analysis and processing module extracts individual feature vectors of children from basic information and behavioral data. The AI ​​analysis and processing module's solution generation unit retrieves expert cases and resources from the correction content library based on the child's individual feature vectors, generates a personalized correction plan, and pushes it to the teacher's terminal in the interactive execution module for teacher review and adjustment. After approval, it is synchronized to the parent's terminal and the child's interactive terminal in the interactive execution module.

[0012] Preferably, step 4 compares behavioral and physiological data before and after correction using an effectiveness evaluation module to calculate a correction effectiveness score. Based on this score, methods for maintaining or optimizing the correction plan include: The correction data is analyzed regularly through the effect evaluation module, and the behavioral and physiological data before and after correction are compared using the data comparison unit in the effect evaluation module. The quantitative scoring unit in the effectiveness evaluation module calculates the corrective effect score based on the comparison results using a percentage-based scoring method. If the score is ≤50, the current correction plan will continue to be implemented; if the score is >50, the strategy adjustment unit in the effect evaluation module will trigger the AI ​​analysis and processing module to optimize the correction plan and push it back to each terminal.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: High degree of personalization: AI algorithms accurately extract individual characteristics of children and generate highly adaptable correction plans, solving the problem of the "one-size-fits-all" approach of traditional methods and greatly improving the targeted nature of correction.

[0014] Significant corrective effect: The introduction of an assessment method that combines physiological monitoring and AI analysis enables dynamic adjustment of the correction process, increasing the success rate by more than 30% compared to traditional methods, and shortening the correction cycle by an average of 20%.

[0015] Close integration of teaching and practice: The course "Behavior Modification" is built into an integrated platform that provides rich real-world cases and data resources, effectively solving the problems of rigid teaching content and disconnect from practice, and improving the quality of course teaching.

[0016] Easy and efficient to operate: Through multi-terminal interactive design, the operation process for teachers and parents is simplified, and automated data collection and analysis reduce manual workload and improve the efficiency of behavior correction work.

[0017] Wide range of applications: It can be applied to the correction of various problem behaviors in normal preschool children and children with special needs (autism, attention deficit, etc.). It can also be used as a teaching and training platform for behavior correction courses, combining practical application and teaching value. Attached Figure Description

[0018] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an architecture diagram of an AI-enabled preschool inclusive education behavior correction system according to an embodiment of the present invention, showing the composition and connection relationship of each module of the system, as well as the connection logic between each module and the terminal device; Figure 2 This is a flowchart of the AI-enabled preschool inclusive education behavior correction method according to an embodiment of the present invention. It shows the complete process from data collection, behavior analysis, plan generation, correction implementation, effect evaluation to teaching feedback, as well as the key operation steps and data flow logic of each stage. Figure 3 This is a schematic diagram of the internal structure of the AI ​​analysis and processing module in an embodiment of the present invention, illustrating the internal workflow and data interaction relationships of the behavior recognition unit, feature extraction unit, and scheme generation unit. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Example 1 like Figure 1 As shown, this invention provides an AI-enabled preschool inclusive education behavior correction system, comprising: Data acquisition module, AI analysis and processing module, correction content library, interactive execution module, physiological monitoring module, effect evaluation module, and teaching feedback module; The data acquisition module is used to collect basic information, behavioral data, and teaching scenario data of children to be corrected; The correction content library is used to store behavioral correction resources related to the field of preschool inclusive education; The AI ​​analysis and processing module is used to identify and extract features from the data collected by the data acquisition module, and generate a correction plan by combining it with behavior correction resources. The interactive execution module is used to enable teachers and parents to guide children to be corrected to participate in corrective training based on the corrective plan; The physiological monitoring module is used to collect physiological data of the child to be corrected during the correction process and transmit it to the effect evaluation module; The effect evaluation module is used to compare and quantify behavioral data, physiological data and baseline data before correction to obtain a correction effect score, and adjust the correction plan based on the correction effect score. The teaching feedback module is used to synchronize data, cases, and assessment results from the correction process to the behavior correction course teaching platform, providing practical case support and data resources for the course.

[0023] The specific implementation process of this invention is as follows: Data Acquisition Module: This module collects basic information, behavioral data, and teaching scenario data for children requiring intervention. Basic information includes age, gender, cognitive level, health status, and family background. Behavioral data is obtained through video surveillance, teacher observation records, and parental feedback, including problem behavior types (such as aggressive behavior, autism-related behaviors, inattention, etc.), frequency, duration, and triggering scenarios. Teaching scenario data includes the classroom environment, types of teaching activities, and teacher-student interaction. The collected data undergoes preprocessing such as noise reduction, standardization, and feature extraction.

[0024] AI Analysis and Processing Module: As the core of the system, it includes a behavior recognition unit, a feature extraction unit, and a solution generation unit. The internal structure of the AI ​​Analysis and Processing Module is as follows: Figure 3 As shown, the behavior recognition unit employs computer vision technology and machine learning algorithms (such as CNN convolutional neural networks and LSTM recurrent neural networks) to automatically identify and classify collected behavioral data, accurately determining the type of problematic behavior. The primary task of this unit is to accurately identify and determine the types of problematic behaviors in preschool children. Its core function is to provide clear targets for subsequent personalized interventions. Only by accurately identifying specific problematic behaviors (such as aggressive behavior, self-harm behavior, and stereotyped behavior) can the system "prescribe the right medicine," calling upon relevant expert knowledge and teaching resources to lay the foundation for generating effective intervention plans. Specifically, the behavior recognition unit constructs a CNN-LSTM hybrid model, which combines the advantages of convolutional neural networks (CNN) in spatial feature extraction with the advantages of long short-term memory networks (LSTM) in time series modeling.

[0025] 1. Input layer: Receives the pre-processed video frame sequence {X1, X2, ..., X...} t}, where X t This represents the RGB image at time t.

[0026] 2. CNN Spatial Feature Extraction Layer: A pre-trained 2D or 3D CNN (such as ResNet-50, I3D) is used as the backbone network. Convolution operations are performed on each frame to extract high-level spatial features. This process can be formally represented as: f t =CNN(X t ;θ cnn ) Among them, f t The spatial feature vector θ is extracted at time t. cnn These are the weight parameters of a CNN network.

[0027] 3. LSTM Temporal Modeling Layer: This layer models the spatial feature vectors {f1, f2, ..., f...} output by the CNN layers. t The input is fed into a multi-layer LSTM network. LSTM utilizes its unique gating mechanism (input gate i) to... t Forgotten Gate t Output gate o t Learn the dynamic changes and dependencies of actions over time.

[0028] Among them, h t It is the hidden state at the current moment (i.e., the output of the LSTM), c t It is the internal memory unit of the LSTM unit at time step t, also known as "cellular state" or "long-term memory". t The input data at the current moment, i t The activation value of the input gate at time t determines the current input x. t How much new information can be written into long-term memory, f t The activation value of the ForgetGate at time t determines the long-term memory c of the previous time step. t−1 How much information can be retained? t The activation value of the output gate at time t determines the current value of c. t How much information can be output to the hidden state h? t . t The current state of a candidate cell is based on the current input x. t and the hidden state h from the previous moment t−1 The calculated "new information to be written". xi In the input gate, the current input x t Connected weight matrices. W hi In the input gate, the hidden state h from the previous time step... t−1Connected weight matrices; W xf It is in the forget gate and the current input x t Connected weight matrices; W hf It is the hidden state h in the forget gate compared to the previous moment. t−1 Connected weight matrices; W xo For the output gate, the current input x t Connected weight matrices; W ho The output gate contains the hidden state h from the previous time step. t−1 Connected weight matrices; W xc For candidate cell states, the current input x t Connected weight matrices; W hc In the candidate cell state, compared with the hidden state h in the previous time step t−1 A contiguous weight matrix. σ is the sigmoid function, tanh is the hyperbolic tangent function, ⊙ denotes element-wise multiplication, and W and b are the learnable weights and biases.

[0029] 4. Classification Output Layer: This layer stores the hidden state h from the last time step of the LSTM layer. t Alternatively, the pooling results from all time steps are input into a fully connected layer, and finally the probability distribution of the behavior belonging to each category is output through the Softmax function.

[0030] Here, ŷ is the predicted probability vector, and the category corresponding to its maximum value is the final determined problem behavior type. Softmax (the activation function) is used to transform a set of arbitrary real numbers into a probability distribution. h T W is the hidden state output by the LSTM network at the last time step (time T). y The weight matrix W of the output layer y Its function is to determine h T Which features are useful for the final classification, and assign different weights to these features. y The bias term of the output layer can be understood as a "baseline score" or a "correction term". Even if W y h T The calculated results were all 0, plus b yThis also allows the model to have an initial output tendency, which can be used to adjust the classification baseline and increase the model's flexibility. Model training is conducted based on the behavioral characteristics of preschool children to improve the accuracy of problem behavior recognition, achieving an accuracy of over 90%. The feature extraction unit extracts individual feature vectors from basic information and behavioral data, including behavioral preferences, emotional response patterns, and learning acceptance abilities. The solution generation unit, based on the feature vectors and combined with expert cases and teaching resources in the correction content library, uses collaborative filtering algorithms and decision tree models, considering multiple dimensions such as the child's individual feature vectors, family environment, and teaching scenarios, to generate personalized correction solutions. This achieves precise matching of correction solutions, including correction goals, specific techniques (such as positive reinforcement, extinction, and behavior shaping), implementation steps, and expected results. The user-based collaborative filtering algorithm treats the current child's individual feature vector as a "user" and calculates its similarity to historical cases (successful cases with similar characteristics recorded by experts) in the correction content library. The similarity calculation uses cosine similarity or Pearson correlation coefficient to calculate the similarity between vectors, sim(u,v). Select the correction schemes used in the K most similar historical cases as the initial candidate scheme set C.

[0031] Here, u and v refer to the two objects whose similarity is being calculated. u represents the child currently being evaluated, and v represents a historical case in the correction content library. sim(u,v) represents the similarity between user u and user v. The higher this value, the more similar the current child u is to the child v in the historical case in terms of behavioral characteristics, and the more worthy of reference the correction plan used in case v is. I refers to the evaluation dimensions or behavioral characteristics that the two users u and v have interacted or rated together. i is a specific item in set I, for example, i could represent the specific feature dimension of "emotional fluctuation amplitude". r u,i This can be compared to the value of user u on feature dimension i. u is the average rating of user uu for all items, ∑ i∈I This sums up all items in set II. A CART (Classification and Regression Tree) decision tree model is used to refine and adapt candidate solutions C. The decision tree takes the child's individual feature vector and multiple factors such as family environment (e.g., parental involvement, parenting style) and teaching scenario (e.g., integrated class size, support resources) as input features. The decision-making process starts from the root node and proceeds downwards along the decision path based on the values ​​of the input features (e.g., "emotional fluctuation amplitude" > threshold), eventually reaching one or more leaf nodes. Each leaf node is associated with one or more specific correction solution components and an estimated "fitness" score. During the construction of the decision tree, the Gini coefficient (Gini Impurity) is used to select the optimal splitting features and split points, ensuring that cases within each child node are as consistent as possible in terms of "correction solution effectiveness."

[0032] Where D is the sample set at the current node, K is the number of categories, and p k is the probability of the k-th class sample appearing in set D, and Gini(D) is the Gini coefficient of dataset D, calculated between [0,1). Finally, the algorithm combines the "common experience" of collaborative filtering with the "personalized adaptation" of decision trees to generate a complete and executable personalized correction plan.

[0033] The Correction Content Library stores behavioral correction resources related to the field of inclusive preschool education, including an expert case library, a correction technique library, and a teaching material library. The expert case library contains successful correction cases of different types of problem behaviors, covering case background, correction process, and effectiveness data; the correction technique library contains detailed explanations, operating procedures, and applicable scenarios for various behavioral correction techniques suitable for preschool children; and the teaching material library includes micro-lesson videos, interactive courseware, simulated training scenarios, parent guides, and other teaching resources.

[0034] The interactive execution module includes teacher terminals, parent terminals, and child interaction terminals. The teacher terminal is used to view child data and correction plans, issue correction tasks, record the implementation process, and adjust correction strategies. The parent terminal receives correction tasks from the home terminal and provides feedback on the child's behavior at home. The child interaction terminal uses a cartoonish and fun interface design, presenting correction content through animation and games, adapting to the cognitive characteristics of preschool children, guiding children to participate in correction training, and collecting real-time interactive feedback data. The main purpose of collecting this real-time interactive feedback data is to provide the most direct and objective input for the next round of analysis and plan iteration in the "AI analysis and processing module." When the "plan generation unit" receives the updated feature vector and evaluation data on the effectiveness of the current plan, it determines whether the current correction plan needs fine-tuning (such as adjusting task difficulty or changing the interaction format), or triggers a larger-scale plan regeneration process if the effect remains unsatisfactory. This forms a closed-loop intelligent system of "plan execution → data feedback → AI analysis → plan optimization," ensuring that the correction strategy can dynamically adapt to the child's developmental changes.

[0035] Physiological monitoring module: This module collects physiological data from children during the correction process, including respiratory data, heart rate data, and skin resistance data, and monitors the child's emotional state and concentration in real time. This module works synchronously with the interactive execution module, continuously collecting physiological data and transmitting it to the effect evaluation module while the child is engaged in interactive training. The normal physiological ranges set for the physiological monitoring module are: respiratory rate 15-25 breaths / minute, heart rate 80-120 beats / minute, and skin resistance 10-100kΩ. The system automatically pauses training and issues a prompt when physiological data exceeds the preset normal range.

[0036] The effectiveness evaluation module includes a data comparison unit, a quantitative scoring unit, and a strategy adjustment unit. The data comparison unit compares behavioral and physiological data during the correction process with baseline data before correction. The quantitative scoring unit constructs a quantitative scoring module for effectiveness evaluation. The model adopts the Analytic Hierarchy Process (AHP) to build a three-layer evaluation index system: Goal layer (overall correction effect) → Criterion layer (degree of behavioral improvement, physiological stability, and interaction participation) → Indicator layer (specific quantitative indicators under each criterion). A percentage-based scoring method is used, with the degree of behavioral improvement (40% weight) calculated using the formula B=100. (1-(P_post / P_pre)), where B represents the degree of behavioral improvement, P_pre is the weekly average frequency of the problem behavior before correction, and P_post is the weekly average frequency of the problem behavior during the assessment period. This indicator reflects the relative improvement rate of the problem behavior. Physiological stability (30% weight) mainly includes heart rate variability (HRV), skin conductance activity (SCL), and sleep quality index.

[0037] Where S is the physiological stability score, HRV post It is the mean heart rate variability after intervention, SCL post It is the average skin conductance activity after intervention, Sleep post The sleep quality index after intervention; HRV opt This refers to the individual's optimal heart rate variability baseline value, SCL. opt The optimal skin activity baseline value for an individual, Sleep opt The optimal sleep quality benchmark is the individual's best stable state value extracted from baseline data (e.g., the mean value corresponding to the period with the highest and least fluctuating HRV). The model calculates the deviation of each indicator from the optimal state, takes the average, and then converts it into a percentage score. The smaller the deviation, the higher the score, indicating a more stable physiological state. Interaction participation (30% weight) is used to calculate the corrective effect score across three dimensions, using the formula E=100. (α C_rate+β A_dur+γ T_rate, where E is the interaction engagement rate, C_rate is the task completion rate (number of completed tasks / total number of assigned tasks); A_dur is the ratio of average attention duration to the preset target duration of the plan (maximum of 1); T_rate is the interaction accuracy rate (number of correct interactive operations / total number of interactions); α, β, and γ are weighting coefficients, satisfying α + β + γ = 1, which can be dynamically adjusted according to the child's age and correction goals (default values ​​are set to α = 0.4, β = 0.3, γ = 0.3). The scores of the three dimensions are weighted and summed to obtain the final comprehensive effect score. The score is calculated as 0.4 × B + 0.3 × S + 0.3 × E. A score of 0-25 indicates no risk (excellent correction effect), 26-50 indicates mild risk (good correction effect), 51-75 indicates moderate risk (average correction effect), and 76-100 indicates high risk (poor correction effect). Based on the assessment results, if the score is 51 or higher, the strategy adjustment unit automatically triggers the AI ​​analysis and processing module to regenerate or optimize the correction plan. A mapping relationship is established between physiological data and emotional state and behavioral performance, improving the scientific rigor of the correction effect assessment through multi-dimensional data fusion.

[0038] Teaching Feedback Module: This module is used to synchronize data, cases, and assessment results from the correction process to the behavior correction course teaching platform through an integrated teaching and practice linkage mechanism. It provides the behavior correction course with real and vivid practical case support and data resources, including typical case analyses that students can view and teaching data reports that teachers can use, so as to achieve a deep integration of teaching and practice.

[0039] In summary, this invention achieves accurate identification and analysis of problem behaviors in preschool children, and formulates personalized correction plans based on individual child characteristics; it introduces AI technology and physiological monitoring methods to improve the scientific rigor and objectivity of the correction process, and overcomes the subjectivity of manual assessment; it builds an integrated teaching and practice platform, providing real-world cases and data support for behavior correction courses, thereby improving teaching effectiveness; and it enables dynamic tracking and quantitative evaluation of correction effects, allowing for timely adjustments to correction strategies and improving correction efficiency.

[0040] Example 2 like Figure 2 As shown, based on the same inventive concept, this invention also provides AI-enabled preschool inclusive education behavior correction, implemented using the system described in the foregoing embodiments, with the method including: Step 1: Collect basic information, problem behavior data, and teaching scenario data of the children to be corrected through the data acquisition module. At the same time, collect baseline physiological data of the children to be corrected before correction through the physiological monitoring module. Step 2: Use the AI ​​analysis and processing module to automatically identify and classify the behavioral data and extract the individual feature vectors of children. Combine this with the behavioral correction resources in the correction content library to generate a correction plan. Step 3: Through the interactive execution module, teachers and parents guide children to participate in corrective training according to the correction plan, and collect interactive feedback data and physiological data simultaneously. Step 4: Compare behavioral and physiological data before and after correction through the effect evaluation module, calculate the correction effect score, and maintain or optimize the correction plan based on the correction effect score; Step 5: Synchronize case data and assessment reports from the correction process to the behavior correction course teaching platform through the teaching feedback module, update the teaching material library, and provide practical support for course teaching; Step 6: Repeat steps 3-5 until the child's correction effect score is ≤25 points for two consecutive times. Once this is completed, the final correction report will be generated and archived in the correction content library and teaching platform.

[0041] The specific implementation process of this invention is as follows: Data collection phase: The data collection module collects basic information, problem behavior data, and teaching scenario data of the children to be corrected, and establishes a personal data file for each child; the physiological monitoring module collects the children's baseline physiological data before correction (normal range of respiration, heart rate, and skin resistance) as a reference for subsequent evaluation.

[0042] Behavior analysis and solution generation stage: The behavior recognition unit of the AI ​​analysis and processing module automatically identifies and classifies the collected behavior data to determine the type of problem behavior; the feature extraction unit extracts individual feature vectors for children; the solution generation unit, based on the feature vectors, retrieves expert cases and resources from the correction content library, generates personalized correction solutions, and pushes them to the teacher's terminal for teacher review and adjustment. After approval, the solutions are synchronized to the parent's terminal and the child's interactive terminal.

[0043] Correction Implementation Phase: Teachers and parents guide children to participate in correction training through the interactive execution module, based on the correction plan. Children engage in fun training through an interactive terminal, and the interactive execution module collects children's interactive feedback data in real time. The physiological monitoring module simultaneously collects children's physiological data during training. If the physiological data exceeds the normal range, the system automatically pauses training and issues a prompt, resuming only after the child's condition stabilizes.

[0044] Effectiveness evaluation phase: The effectiveness evaluation module analyzes the correction data regularly (once a week). The data comparison unit compares the behavioral and physiological data before and after correction. The quantitative scoring unit calculates the correction effect score. If the score is ≤50 points, the current correction plan is maintained and continues to be implemented. If the score is >50 points, the strategy adjustment unit triggers the AI ​​analysis and processing module to optimize the correction plan and push it back to each terminal.

[0045] Teaching feedback phase: The teaching feedback module synchronizes case data and assessment reports from the correction process to the behavior correction course teaching platform, updates the teaching material library, and provides practical support for course teaching.

[0046] Repeated execution phase: Repeat steps 3-5 until the child's correction effect score is ≤25 points for two consecutive times. The correction is then considered complete, a final correction report is generated, and the report is archived in the correction content library and teaching platform.

[0047] Example 3 This embodiment illustrates the specific implementation process of the system described in the foregoing embodiment using specific data: (1) System hardware configuration Data acquisition equipment: a high-definition camera (1080P resolution, 30fps) for capturing videos of children's behavior; a smart bracelet (supporting heart rate, respiration, and skin resistance monitoring, with a sampling frequency of 1Hz) for collecting physiological data; and a tablet computer (Android 11 or above or iOS 14 or above, 4GB+ RAM and 64GB+ storage) as a teacher's terminal, a parent's terminal, and a child's interactive terminal.

[0048] Servers: Cloud servers (8 CPU cores, 16GB memory, 500GB storage, supporting GPU acceleration) are used to run AI algorithms, store data, and manage system services.

[0049] Network environment: Supports WiFi 6 or wired network with bandwidth ≥100Mbps to ensure real-time data transmission.

[0050] (2) Software implementation details Data acquisition module: Video data is acquired through a camera, and frame extraction and preprocessing are performed using OpenCV; the smart bracelet connects to the terminal device via Bluetooth 5.0 to transmit physiological data in real time; teachers and parents manually enter supplementary data through the terminal APP, and the data format is uniformly JSON.

[0051] AI analysis and processing module: Developed using Python, it uses the TensorFlow framework to build a CNN-LSTM hybrid model. The model training dataset contains 1000+ videos of various behaviors of preschool children and their corresponding labels. After training, the model's recognition accuracy reaches 92%. The collaborative filtering algorithm is optimized using the biasSVD algorithm to improve the accuracy of the recommended solutions. The decision tree model uses the C4.5 algorithm and builds decision rules based on an expert case library.

[0052] Correction Content Library: Data is stored using a MySQL database. The expert case library contains 500+ successful correction cases, the correction technology library covers 20+ behavior correction techniques applicable to preschool children, and the teaching material library contains 100+ micro-lesson videos, 50+ interactive courseware, and 30+ simulated training scenarios.

[0053] Interactive execution module: The terminal APP is developed using the Flutter framework and supports cross-platform operation; the children's interactive terminal interface adopts a cartoon design, incorporating elements such as animation and games, which is in line with the cognitive characteristics of preschool children; data from each terminal is synchronized with the server through a RESTful API interface, and data transmission is encrypted with AES to ensure data security.

[0054] Physiological monitoring module: Collects respiratory, heart rate, and skin resistance data through smart bracelet, sets normal physiological range (respiratory rate 15-25 breaths / minute, heart rate 80-120 beats / minute, skin resistance 10-100kΩ), and issues audible and visual alerts when the range is exceeded.

[0055] The effectiveness evaluation module uses Python to perform data comparison and quantitative scoring, and generates evaluation reports using Excel. The strategy adjustment unit calls the AI ​​analysis and processing module through an API interface to automatically optimize the solution.

[0056] Teaching feedback module: It interfaces with the school's teaching platform to synchronize case data and evaluation reports, allowing teachers and students to view and download them online.

[0057] (3) Specific implementation cases Taking the correction of insufficient social interaction in a 5-year-old autistic child as an example, the specific implementation process is as follows: Data collection: The video of the child's behavior in the kindergarten classroom and activity area is collected by camera, and the baseline physiological data before correction is collected by smart bracelet (respiratory rate 20 breaths / minute, heart rate 95 beats / minute, skin resistance 50kΩ); the teacher enters the child's basic information (age 5 years old, male, autism spectrum disorder, moderate cognitive level), and the parents provide feedback on the family environment and the child's behavior at home.

[0058] Behavior Analysis and Solution Generation: The behavior recognition unit of the AI ​​analysis and processing module identifies problematic behaviors in children, such as social interaction avoidance and limited verbal responses, from the videos; the feature extraction unit extracts individual characteristics of children (strong interest in animation, stable emotions, and short-term attention span); the solution generation unit searches the correction content library to generate personalized correction plans: with "animation guidance + social simulation + positive reinforcement" as the core, children learn about animated social scenarios for 15 minutes daily through interactive terminals, teachers arrange 5 minutes of one-on-one social interaction training in the classroom, and parents conduct 10 minutes of parent-child interactive games at home, using stickers and verbal praise as reinforcements.

[0059] Correction Implementation: Children watch social-themed animations through interactive terminals and participate in interactive tasks within the animations; teachers guide children to engage in simple interactions with peers in the classroom according to the plan (such as greeting each other and sharing toys), and record the interactions; parents conduct parent-child games at home and provide feedback on the children's participation; the physiological monitoring module collects children's physiological data in real time during training, and no abnormalities were found.

[0060] Effectiveness Evaluation: After the first week, the effectiveness evaluation module compared the data and found that the child's social interaction frequency increased by 3 times / day compared to before the correction, the heart rate stabilized at 90-100 beats / minute, the interaction participation was good, and the quantitative score was 45 points (mild risk, good correction effect). The original plan was maintained and continued. After the third week, the child's social interaction frequency increased by 8 times / day, and the child was able to initiate simple communication on his own initiative, with a quantitative score of 22 points (no risk, excellent correction effect).

[0061] Teaching feedback: The system synchronizes the correction process, data, and assessment report of this case to the teaching platform of the "Behavior Modification" course, so that it can be used as a practical case for students to learn and analyze.

[0062] Correction complete: After the fourth week, if the child's correction effect score is 20 points, and ≤25 points for two consecutive times, the system will determine that the correction is complete, generate a final correction report and archive it.

[0063] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. An AI-enabled preschool inclusive education behavior correction system, characterized in that, The system includes: a data acquisition module, an AI analysis and processing module, a correction content library, an interactive execution module, a physiological monitoring module, an effect evaluation module, and a teaching feedback module; The data acquisition module is used to collect basic information, behavioral data, and teaching scenario data of children to be corrected; The correction content library is used to store behavioral correction resources related to the field of preschool inclusive education; The AI ​​analysis and processing module is used to identify and extract features from the data collected by the data acquisition module, and generate a correction plan by combining it with behavior correction resources. The interactive execution module is used to enable teachers and parents to guide children to be corrected to participate in corrective training based on the corrective plan; The physiological monitoring module is used to collect physiological data of the child to be corrected during the correction process and transmit it to the effect evaluation module; The effect evaluation module is used to compare and quantify behavioral data, physiological data and baseline data before correction to obtain a correction effect score, and adjust the correction plan based on the correction effect score. The teaching feedback module is used to synchronize data, cases, and assessment results from the correction process to the behavior correction course teaching platform, providing practical case support and data resources for the course.

2. The system according to claim 1, characterized in that, The AI ​​analysis and processing module includes: a behavior recognition unit, a feature extraction unit, and a solution generation unit; The behavior recognition unit is used to automatically identify and classify behavioral data using computer vision technology and machine learning algorithms, and accurately determine the type of problematic behavior. The feature extraction unit is used to extract individual feature vectors of children from basic information and behavioral data, including behavioral preferences, emotional response patterns, and learning acceptance abilities; The solution generation unit is used to generate personalized correction solutions based on the child's individual feature vectors, combined with expert cases and resources in the correction content library, through collaborative filtering algorithms and decision tree models.

3. The system according to claim 1, characterized in that, The interactive execution module includes: teacher terminal, parent terminal, and children's interactive terminal; The teacher terminal is used to view children's data and correction plans, issue correction tasks, record the correction implementation process, and adjust correction strategies. Parent terminal is used to receive correction tasks from the home terminal and provide feedback on the child's behavior at home; The children's interactive terminal uses a fun interface design to present corrective content, guide children to participate in corrective training, and collect children's interactive feedback data in real time.

4. The system according to claim 1, characterized in that, The effectiveness evaluation module includes: a data comparison unit, a quantitative scoring unit, and a strategy adjustment unit; The data comparison unit is used to compare behavioral and physiological data during the correction process with baseline data before correction. The quantitative scoring unit uses a percentage-based scoring method to calculate the correction effect score from three dimensions: degree of behavioral improvement, physiological stability, and interaction participation, based on the comparison results. The strategy adjustment unit automatically triggers the AI ​​analysis and processing module to regenerate or optimize the correction plan if the score is greater than or equal to a preset threshold, based on the correction effect score.

5. A preschool inclusive education behavior correction method based on AI empowerment, wherein the method is implemented using the system described in any one of claims 1-4, characterized in that, The method includes: Step 1: Collect basic information, problem behavior data, and teaching scenario data of the children to be corrected through the data acquisition module. At the same time, collect baseline physiological data of the children to be corrected before correction through the physiological monitoring module. Step 2: Use the AI ​​analysis and processing module to automatically identify and classify the behavioral data and extract the individual feature vectors of children. Combine this with the behavioral correction resources in the correction content library to generate a correction plan. Step 3: Through the interactive execution module, teachers and parents guide children to participate in corrective training according to the correction plan, and collect interactive feedback data and physiological data simultaneously. Step 4: Compare behavioral and physiological data before and after correction through the effect evaluation module, calculate the correction effect score, and maintain or optimize the correction plan based on the correction effect score; Step 5: Synchronize case data and assessment reports from the correction process to the behavior correction course teaching platform through the teaching feedback module, update the teaching material library, and provide practical support for course teaching; Step 6: Repeat steps 3-5 until the child's correction effect score is ≤25 points for two consecutive times. Once this is completed, the final correction report will be generated and archived in the correction content library and teaching platform.

6. The method according to claim 5, characterized in that, Step 2 uses an AI analysis and processing module to automatically identify and classify behavioral data and extract individual child feature vectors. Combined with behavioral correction resources in the correction content library, the method for generating a correction plan includes: The behavior recognition unit of the AI ​​analysis and processing module automatically identifies and classifies behavioral data to determine the type of problematic behavior. The feature extraction unit of the AI ​​analysis and processing module extracts individual feature vectors of children from basic information and behavioral data. The AI ​​analysis and processing module's solution generation unit retrieves expert cases and resources from the correction content library based on the child's individual feature vectors, generates a personalized correction plan, and pushes it to the teacher's terminal in the interactive execution module for teacher review and adjustment. After approval, it is synchronized to the parent's terminal and the child's interactive terminal in the interactive execution module.

7. The method according to claim 5, characterized in that, Step 4 compares behavioral and physiological data before and after correction using the effect evaluation module, calculates the correction effect score, and, based on the correction effect score, methods for maintaining or optimizing the correction plan include: The correction data is analyzed regularly through the effect evaluation module, and the behavioral and physiological data before and after correction are compared using the data comparison unit in the effect evaluation module. The quantitative scoring unit in the effectiveness evaluation module calculates the corrective effect score based on the comparison results using a percentage-based scoring method. If the score is ≤50, the current correction plan will continue to be implemented; if the score is >50, the strategy adjustment unit in the effect evaluation module will trigger the AI ​​analysis and processing module to optimize the correction plan and push it back to each terminal.

Images