Infant teacher practical training supervision system and method

Abstract

The invention provides an infant teacher practical training supervision system and method, and relates to the field of education technology and artificial intelligence cross technology, and the system comprises an AR scene simulation module which is used for constructing a virtual infant teaching scene with a self-adaptive teaching scene; the multi-modal sensing array module is used for synchronously collecting multi-modal teaching characteristics of a teacher; the teaching ability evaluation module is used for evaluating the teaching ability of the teacher through a preset teaching behavior evaluation model based on the multi-modal teaching characteristics of the teacher and the real-time feedback data of the virtual child roles; and the intelligent intervention module is used for generating AR visual prompt information containing specific improvement measures according to the teaching ability evaluation result of the teacher, and displaying the AR visual prompt information in the virtual teaching scene in real time through the AR scene simulation module. According to the invention, the problems of lack of real-time guidance and insufficient interactive feedback in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of educational technology and artificial intelligence, specifically to a preschool teacher training and supervision system and method that integrates augmented reality (AR) scenario simulation, multimodal behavior perception and real-time intervention feedback, and is particularly suitable for in-service training of preschool teachers and training to improve their teaching practice abilities. Background Technology

[0002] In the field of early childhood teacher training, practical training plays a crucial role in enhancing teaching practice abilities. Traditional teacher training mainly relies on theoretical lectures combined with limited teaching internships. This model makes it difficult for trainees to fully master core skills unique to early childhood education, such as non-verbal communication and emotional guidance. With the development of educational technology, immersive simulation training has become an important way to bridge the gap between theory and practice. Among these, augmented reality (AR) technology, due to its virtual-real integration characteristics, is particularly suitable for constructing safe, controllable, and highly realistic early childhood teaching scenarios.

[0003] In the existing technology, patent CN120894193A, entitled "A Method and System for Cultivating Teachers' Classroom Training Ability Based on Real-Time Interaction with Virtual Students," proposes a teacher training system based on virtual students. However, its main drawback lies in the unidirectional nature of its interaction design. This system uses a pre-set script with a fixed process to drive virtual student behavior, failing to simulate the random interactive events unique to early childhood classrooms (such as sudden emotional fluctuations or unconventional questions). Its evaluation mechanism relies on video playback analysis after training, lacking real-time feedback and dynamic guidance on teachers' teaching behavior, making it difficult for trainees to correct inappropriate teaching strategies in a timely manner. This lagging evaluation model is significantly different from the immediate interactive response characteristics required in early childhood education, hindering the improvement of training effectiveness. Summary of the Invention

[0004] In view of this, the present invention provides a training supervision system and method for preschool teachers, which aims to solve the problems of lack of real-time guidance and insufficient interactive feedback in the existing technology.

[0005] Therefore, the present invention provides the following technical solution:

[0006] On the one hand, the present invention provides a preschool teacher training supervision system, comprising:

[0007] The AR scenario simulation module is used to construct virtual early childhood teaching scenarios with adaptive teaching contexts; the virtual early childhood teaching scenarios can dynamically adjust the interactive feedback of the virtual early childhood characters according to the teacher's teaching behavior.

[0008] A multimodal perception array module is used to synchronously collect the teacher's multimodal teaching features; the multimodal teaching features include: facial expressions, body movements, voice features, and physiological features;

[0009] The teaching ability assessment module is connected to the AR scenario simulation module and the multimodal perception array module, respectively. It is used to assess the teacher's teaching ability based on the teacher's multimodal teaching characteristics and real-time feedback data of the virtual child role, through a preset teaching behavior assessment model.

[0010] The intelligent intervention module is connected to both the AR scenario simulation module and the teaching ability assessment module. It is used to generate AR visual prompts containing specific improvement measures based on the teacher's teaching ability assessment results, and to display them in real time in the virtual teaching scenario through the AR scenario simulation module.

[0011] Furthermore, the preschool teacher training supervision system also includes a feedback record and growth tracking database, which is connected to the AR scenario simulation module, the multimodal perception array module, the teaching ability assessment module, and the intelligent intervention module, respectively, to store data during the teacher training process and generate teacher ability development reports through data analysis.

[0012] Furthermore, the AR scenario simulation module includes:

[0013] The scene construction submodule is used to build interactive virtual teaching scenes for young children;

[0014] The role behavior library stores feedback behavior patterns of children of different ages in response to various teaching behaviors of teachers;

[0015] The interactive response submodule is used to match the feedback behavior pattern of the virtual child role from the role behavior library based on the teacher's teaching behavior, and trigger the feedback behavior pattern of the virtual child role.

[0016] The AR visualization prompt submodule is used to display AR visualization prompts containing specific improvement measures in real time in virtual teaching scenarios.

[0017] Furthermore, the multimodal sensing array module includes:

[0018] Multiple high-definition cameras are used to capture teachers' facial expressions and body movements;

[0019] A depth sensor that provides three-dimensional motion trajectory data;

[0020] A microphone array for collecting voice signals; and wearable biosensors for monitoring teachers' physiological indicators.

[0021] Furthermore, the teaching behavior assessment model is implemented using a multi-layer neural network architecture, including:

[0022] Input layer: Used to receive teacher teaching characteristics and virtual child feedback data;

[0023] Feature extraction layer: Contains multiple parallel convolutional neural network subnetworks, each processing teaching features of different modalities;

[0024] Temporal modeling layer: Employs a long short-term memory temporal network to capture the dynamic temporal characteristics of teaching behavior;

[0025] Decision fusion layer: Weighted fusion of multimodal temporal features through an attention mechanism;

[0026] Evaluation output layer: A softmax classifier is used to map the fused features into a multi-dimensional evaluation result of teaching ability. The evaluation dimensions include the appropriateness of teaching strategies, the quality of teacher-student interaction, emotion management ability, and teaching innovation.

[0027] Furthermore, the intelligent intervention module includes:

[0028] A strategy library stores improvement measures for different teaching problems;

[0029] The improvement measures determination submodule identifies improvement measures corresponding to the teaching problems from the strategy library when the teacher's teaching ability assessment result indicates the existence of teaching problems.

[0030] The AR rendering submodule transforms the improvement measures into visual prompts in the virtual teaching scene and transmits them to the AR scenario simulation module.

[0031] Furthermore, the AR scenario simulation module dynamically adjusts the intensity of visual prompts based on the teacher's improvement progress.

[0032] Furthermore, this invention also provides a method for supervising the practical training of preschool teachers, comprising the following steps:

[0033] A virtual preschool teaching scenario with adaptive teaching context is constructed, wherein the virtual preschool teaching scenario can dynamically adjust the interactive feedback of the virtual preschool character according to the teacher's teaching behavior;

[0034] The system collects teachers' multimodal teaching characteristics in real time, including facial expressions, body movements, voice features, and physiological features.

[0035] Based on the multimodal teaching characteristics of teachers and real-time feedback data from virtual child roles, the teaching ability of teachers is evaluated through a pre-set teaching behavior assessment model.

[0036] Based on the evaluation results of teachers' teaching abilities, AR visual prompts containing specific improvement measures are generated and displayed in real time in the virtual teaching scenario.

[0037] Furthermore, the method for supervising the practical training of preschool teachers also includes the following steps: storing data during the teacher's practical training process and generating a teacher competency development report through data analysis.

[0038] Furthermore, the teaching behavior assessment model is trained using an end-to-end supervised learning framework, including:

[0039] We collected a dataset of teaching behaviors from outstanding historical teachers, and labeled facial expressions, body movements, voice features, and physiological signals to form a sample of teaching behavior features with time-series labels.

[0040] The input multimodal teaching features and virtual preschool feedback data are normalized and time-series aligned.

[0041] Spatial features of each modality are extracted by parallel CNN sub-networks, cross-modal temporal dependencies are captured by LSTM layers, and multimodal feature weights are dynamically fused using an attention mechanism.

[0042] During training, the real labels of the teacher evaluation dimension are used as supervision signals. The network parameters are optimized by the cross-entropy loss function. The weights of the CNN-LSTM-attention module are iteratively adjusted by the backpropagation algorithm. An early stopping strategy is adopted to prevent overfitting, so that the model can automatically learn the mapping relationship of teaching effectiveness from the temporal correlation between teacher behavior and virtual children's feedback.

[0043] Advantages and positive effects of the present invention: The preschool teacher training supervision system and method of the present invention constructs a high-fidelity preschool teaching scene in an AR environment and integrates a multimodal perception module to collect teachers' multimodal teaching characteristics in real time; it combines an intelligent assessment model to assess teachers' teaching abilities and generates immediate intervention instructions based on the assessment results; finally, it guides teachers to adjust their teaching behavior through AR visualization prompts, forming a closed-loop training mechanism of "perception-assessment-intervention", thereby realizing a paradigm shift from passive assessment to active training and significantly improving the pertinence and effectiveness of preschool teacher training. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a structural block diagram of a preschool teacher training supervision system according to an embodiment of the present invention;

[0046] Figure 2This is a structural block diagram of the AR scenario simulation module in an embodiment of the present invention;

[0047] Figure 3 This is a structural block diagram of the intelligent intervention module in an embodiment of the present invention;

[0048] Figure 4 This is a flowchart of a method for supervising the practical training of preschool teachers in an embodiment of the present invention. Detailed Implementation

[0049] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0050] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0051] like Figure 1 As shown in the figure, an embodiment of the present invention provides a preschool teacher training and supervision system, comprising: an AR scenario simulation module 101, a multimodal perception array module 102, a teaching ability assessment module 103, and an intelligent intervention module 104; the teaching ability assessment module 103 is connected to both the AR scenario simulation module 101 and the multimodal perception array module 102; the intelligent intervention module 104 is connected to both the AR scenario simulation module 101 and the teaching ability assessment module 103. Wherein:

[0052] AR scenario simulation module 101 is used to construct a virtual early childhood teaching scenario with adaptive teaching situations. This virtual early childhood teaching scenario can dynamically adjust the interactive feedback of the virtual child characters based on the teacher's teaching behavior. This module uses augmented reality technology, through a head-mounted display device or a projection-based AR system, to construct a highly realistic virtual early childhood teaching scenario for trained teachers. The virtual scenario contains multiple virtual child characters with different personality traits. These characters are constructed based on a model of early childhood behavioral psychology and can generate interactive feedback in accordance with the teacher's teaching behavior and the cognitive characteristics of young children. In particular, the virtual scenario has adaptive characteristics, dynamically adjusting the reaction patterns of the virtual child characters according to the teacher's teaching behavior. For example, when the teacher uses an overly harsh tone, the virtual child characters will show fear or resistance; while when the teacher uses encouraging language, the virtual child characters will show a positive learning attitude. This dynamic adjustment mechanism makes the training environment closer to real teaching situations. Specifically, such as... Figure 2 As shown, the AR scenario simulation module 101 includes: a scenario construction submodule 1011, used to construct interactive virtual teaching scenarios for preschoolers; a role behavior library 1012, storing feedback behavior patterns of preschoolers of different ages in response to various teaching behaviors of teachers; an interactive response submodule 1013, used to match the feedback behavior patterns of virtual preschool characters from the role behavior library 1012 based on the teacher's teaching behaviors, and trigger the feedback behavior patterns of the virtual preschool characters; and an AR visualization prompt submodule 1014, used to display AR visualization prompts containing specific improvement measures in real time within the virtual teaching scenario. In practical applications, the AR scenario simulation module 101 can simulate various typical teaching scenarios, such as group teaching activities, learning area activities, and transitional activities. Each scenario has specific assessment focuses to ensure that teachers receive comprehensive skills training.

[0053] A multimodal sensing array module 102 is used to synchronously collect the teacher's multimodal teaching characteristics, including facial expressions, body movements, voice features, and physiological characteristics. This module includes multiple distributed sensing devices, including multiple sets of high-definition cameras, depth sensors, microphone arrays, and wearable biosensors. The camera array captures the teacher's facial expressions and body movements, the depth sensor provides three-dimensional motion trajectory data, the microphone array collects voice signals, and the biosensors monitor the teacher's physiological indicators such as heart rate and skin conductance. Furthermore, multiple infrared markers can be deployed in the training area, combined with an inertial measurement unit, to achieve millimeter-level precision tracking of the teacher's body movements. This hybrid tracking technology overcomes the shortcomings of simple visual tracking, which is easily affected by occlusion, ensuring the continuity of data acquisition. After time-synchronized processing, these sensor data form a complete dataset of teacher teaching characteristics. In particular, this system employs an advanced sensor fusion algorithm to precisely align data from different modalities in the spatiotemporal dimensions, ensuring the accuracy of subsequent analysis.

[0054] The teaching ability assessment module 103 is used to assess teachers' teaching abilities based on their multimodal teaching characteristics and real-time feedback data from virtual child roles, through a pre-set teaching behavior assessment model. The teaching behavior assessment model is implemented using a multi-layer neural network architecture, including:

[0055] Input layer: Receives teacher teaching characteristics (facial expressions, body movements, voice features, and physiological characteristics) and virtual child feedback data;

[0056] Feature extraction layer: Contains multiple parallel convolutional neural network (CNN) subnetworks, each processing teaching features of different modalities;

[0057] Temporal modeling layer: Employs a Long Short-Term Memory (LSTM) network to capture the dynamic temporal characteristics of teaching behavior;

[0058] Decision fusion layer: Weighted fusion of multimodal temporal features through an attention mechanism;

[0059] The evaluation output layer employs a softmax classifier to map the fused features into multi-dimensional evaluation results. Evaluation dimensions include the appropriateness of teaching strategies, the quality of teacher-student interaction (including sub-indicators such as questioning skills and timely feedback), emotional management ability, and teaching innovation. The system establishes an objective chain of evidence for the effectiveness of teaching behaviors by analyzing the temporal correlation between teacher behavior and virtual children's feedback (e.g., the correspondence between changes in facial expressions and changes in attention).

[0060] The training process of this model adopts an end-to-end supervised learning framework: It collects a dataset of teaching behaviors from historically successful teachers, annotating facial expressions, body movements, speech features, and physiological signals to form temporally labeled teaching behavior feature samples; it normalizes and temporally aligns the input multimodal teaching features (facial expressions, body movements, speech, and physiological signals) and virtual child feedback data; subsequently, it extracts spatial features of each modality through parallel CNN sub-networks, captures cross-modal temporal dependencies using LSTM layers, and dynamically fuses multimodal feature weights using an attention mechanism; during training, it uses real labels from teacher evaluation dimensions (such as strategy appropriateness and interaction quality) as supervision signals, optimizes network parameters using the cross-entropy loss function, iteratively adjusts the weights of the CNN-LSTM-attention module using the backpropagation algorithm, and employs an early stopping strategy to prevent overfitting. Ultimately, the model can automatically learn the mapping relationship of teaching effectiveness from the temporal correlation between teacher behavior and virtual child feedback.

[0061] The intelligent intervention module 104 generates AR visual prompts containing specific improvement measures based on the teacher's teaching ability assessment results, and displays them in real time in a virtual teaching scenario through the AR scenario simulation module 101. This module generates targeted improvement suggestions based on the output of the teaching ability assessment module, and presents them to the teacher visually through the AR scenario simulation module. Specifically, such as... Figure 3 As shown, the intelligent intervention module 104 includes: a strategy library 1041, storing improvement measures for different teaching problems; an improvement measure determination submodule 1042, which identifies improvement measures corresponding to teaching problems from the strategy library 1041 when the teacher's teaching ability assessment result indicates the existence of teaching problems; and an AR rendering submodule 1043, which transforms the improvement measures into visual prompts in a virtual teaching scene and transmits them to the AR scenario simulation module 101. Preferably, when providing visual prompts, the intensity of the prompts can be dynamically adjusted according to the teacher's improvement progress: for recurring problems, the prompts become more prominent; for aspects that have been improved, the prompts gradually fade out. This adaptive intervention mechanism conforms to the learning curve and can effectively promote the improvement of teaching skills.

[0062] In another embodiment, the preschool teacher training supervision system further includes a feedback record and growth tracking database (not shown in the figure), which is connected to the AR scenario simulation module 101, the multimodal perception array module 102, the teaching ability assessment module 103 and the intelligent intervention module 104 respectively, for storing data during the teacher training process and generating teacher ability development reports through data analysis.

[0063] In practical application, the system operates as follows: First, teachers wear AR display devices and biosensors and enter the designated training area. After system initialization, the AR scenario simulation module 101 constructs a virtual classroom environment, and virtual child characters appear in the scene according to preset teaching scenarios. After the teacher begins teaching training, the multimodal perception array module synchronously collects various teaching behavior data. The teaching ability assessment module 103 analyzes this data in real time and correlates it with the feedback data of the virtual children. Based on the analysis results, the intelligent intervention module 104 provides immediate guidance through the AR interface. The entire training process is fully recorded and can be used for subsequent review and analysis.

[0064] The system supports a multi-person collaborative training mode. Multiple teachers can simultaneously enter the same virtual classroom scenario, taking on the roles of lead instructor and assistant instructor, respectively. The system can differentiate the behavioral data of different teachers and provide targeted assessments and guidance. This mode is particularly suitable for cultivating teachers' teamwork skills.

[0065] The preschool teacher training supervision system in the above embodiments enables comprehensive, objective, and real-time evaluation and guidance of teachers' teaching abilities, providing strong technical support for the professional development of preschool teachers.

[0066] like Figure 4 As shown, based on the aforementioned preschool teacher training supervision system, this embodiment provides a preschool teacher training supervision method, including the following steps:

[0067] S1. Construct a virtual preschool teaching scenario with adaptive teaching context, wherein the virtual preschool teaching scenario can dynamically adjust the interactive feedback of the virtual preschool character according to the teacher's teaching behavior;

[0068] In step S1, the virtual teaching scenario construction begins with the system loading corresponding scenario templates based on the training objectives. These templates cover typical scenarios in kindergarten education, such as group teaching activities, corner play guidance, and daily life organization. Each scenario template includes environmental setting parameters, virtual child character configurations, and a description of the teaching task. The system uses an augmented reality engine to render a three-dimensional virtual environment in real time, presenting it to the teacher through a head-mounted display or holographic projection device. The virtual child characters are constructed based on a child behavior psychology model and have age-specific behavioral patterns. Specifically, the implementation process of the virtual child characters is as follows: First, a database of typical behavioral characteristics of children at different age stages is established based on developmental psychology theory, including dimensions such as attention span, language expression level, and social interaction style. Then, a role behavior engine is constructed using a combination of rule-based behavior trees and machine learning. The rule part sets basic response patterns based on child behavior observation scales (such as CBRS), while the machine learning part generates situational adaptive responses through training with a large number of real teacher-child interaction videos. Each virtual character is given personalized personality parameters (such as introversion / extroversion, sensitivity, etc.), which affect the intensity and manner of its response to teacher behavior. After receiving the teacher's teaching behavior input in real time, the system selects the most matching response combination from the preset role behavior library, and finally presents age-characteristic facial expressions, language and action feedback through a 3D animation engine. For example, 3-4 year old characters will show more frequent attention shifts and simple sentence responses, while 5-6 year old characters can show more sustained focus and more complex language interactions.

[0069] S2. Real-time synchronous collection of teachers' multimodal teaching characteristics, wherein the multimodal teaching characteristics include: facial expressions, body movements, voice features and physiological features;

[0070] Step S2, the multimodal feature acquisition process, is implemented using a distributed sensor network. When the teacher conducts teaching activities in a virtual environment, multiple sets of high-definition cameras deployed in the training space capture their facial expressions and body movements from different angles, depth sensors record their spatial movement trajectories, and a microphone array collects speech signals. Simultaneously, biosensors worn by the teacher (such as smart bracelets) monitor physiological indicators such as heart rate and skin conductance. All sensor data are aligned through a time synchronization module to ensure temporal consistency across different modalities. Facial expression analysis employs a deep learning-based facial motion coding system to identify the teacher's emotional state and attention direction; body movement analysis extracts key joint motion features using skeletal tracking technology; and speech analysis combines voiceprint recognition and semantic analysis techniques to assess the teacher's language expression ability. These feature data, after preprocessing, form a standardized multimodal feature vector sequence for subsequent analysis.

[0071] S3. Based on the multimodal teaching characteristics of teachers and real-time feedback data from virtual child roles, the teaching ability of teachers is evaluated through a pre-set teaching behavior evaluation model.

[0072] Step S3, the teaching ability assessment process, is the core analytical component of this method. The assessment model employs a multi-layered neural network architecture, including an input layer, a feature extraction layer, a temporal modeling layer, a decision fusion layer, and an assessment output layer. The feature extraction layer comprises multiple parallel CNN subnetworks, processing facial expressions, body movements, speech features, and physiological features respectively. The temporal modeling layer uses LSTM to capture the dynamic evolution of teaching behaviors. The decision fusion layer integrates the temporal features of each modality through an attention mechanism, ultimately outputting a multi-dimensional assessment result. Assessment dimensions include, but are not limited to: appropriateness of teaching strategies, quality of teacher-student interaction, emotion management ability, and teaching innovation. Each dimension is subdivided into several specific indicators; for example, the quality of teacher-student interaction can be further decomposed into sub-indicators such as questioning skills, timely feedback, and individual attention. During the assessment process, the system compares the correlation between teacher behavior and virtual children's feedback in real time, such as analyzing changes in children's attention after specific changes in the teacher's facial expressions, thereby establishing an objective chain of evidence for the effectiveness of teaching behaviors.

[0073] S4. Generate AR visual prompts containing specific improvement measures based on the teacher's teaching ability assessment results, and display them in real time in the virtual teaching scene;

[0074] Step S4, the intelligent intervention phase, utilizes augmented reality (AR) technology for real-time guidance. The system generates tiered intervention strategies based on the assessment results: for basic teaching standard issues, such as excessively fast speaking speed or insufficient eye contact, immediate prompts are provided via AR floating tags; for complex teaching strategy issues, such as simplistic questioning or abrupt activity transitions, a virtual tutor provides demonstrations and explanations after the teaching segment concludes. The intervention content is designed following a three-part structure: "describe the phenomenon - analyze the impact - suggest improvements," ensuring the guidance's relevance and operability. Notably, the system supports adaptive adjustment of intervention intensity, increasing the prominence of prompts for recurring problems and gradually reducing the frequency of prompts for areas that have improved.

[0075] S5. Store data from the teacher training process and generate teacher competency development reports through data analysis.

[0076] Step S5's data storage and analysis functions provide long-term support for teacher professional development. The system establishes a digital growth profile for each teacher, comprehensively recording multimodal raw data, assessment results, and intervention records from each training session. This data is processed by a learning analytics engine to generate a visualized competency development map, showcasing the progress trajectory of each teacher's teaching abilities. The system also supports group comparison analysis, helping teachers understand their relative position within their peer groups. Predictive models based on historical data can identify teacher development bottlenecks and recommend personalized continuing education content. The data reporting function can generate standardized assessment reports that meet professional certification requirements, providing objective evidence for teachers' professional development.

[0077] Furthermore, this method supports a collaborative training model. Multiple teachers can simultaneously enter the virtual environment, each assuming different roles such as lead lecturer or assistant lecturer. The system can differentiate the behavioral data of each teacher, providing role suitability assessments and team collaboration guidance to cultivate teachers' collaborative teaching abilities.

[0078] The aforementioned methods for supervising the practical training of preschool teachers enable precise characterization of teachers' teaching behaviors, scientific assessment of teaching abilities, and intelligent guidance for professional development, providing an effective technical solution for improving the overall quality of the preschool teacher workforce.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A preschool teacher training supervision system, characterized in that, include: AR scenario simulation module (101) is used to construct virtual early childhood teaching scenarios with adaptive teaching situations; The virtual early childhood teaching scenario can dynamically adjust the interactive feedback of the virtual child character based on the teacher's teaching behavior. The multimodal sensing array module (102) is used to synchronously collect the multimodal teaching features of teachers; The multimodal teaching features include: facial expressions, body movements, voice features, and physiological features; The teaching ability assessment module (103) is connected to the AR scenario simulation module (101) and the multimodal perception array module (102) respectively. It is used to assess the teaching ability of the teacher based on the teacher's multimodal teaching characteristics and the real-time feedback data of the virtual child role through a preset teaching behavior assessment model. The intelligent intervention module (104) is connected to the AR scenario simulation module (101) and the teaching ability assessment module (103) respectively. It is used to generate AR visual prompts containing specific improvement measures based on the teaching ability assessment results of the teachers, and to display them in real time in the virtual teaching scene through the AR scenario simulation module (101).

2. The preschool teacher training supervision system according to claim 1, characterized in that, Also includes: The feedback record and growth tracking database are connected to the AR scenario simulation module (101), the multimodal perception array module (102), the teaching ability assessment module (103), and the intelligent intervention module (104), respectively, to store data during the teacher training process and generate teacher ability development reports through data analysis.

3. The preschool teacher training supervision system according to claim 1, characterized in that, The AR scenario simulation module (101) includes: The scene construction submodule (1011) is used to build interactive virtual teaching scenes for young children; The role behavior library (1012) stores the feedback behavior patterns of children of different ages in response to various teaching behaviors of teachers; The interactive response submodule (1013) is used to match the feedback behavior pattern of the virtual child role from the role behavior library (1012) according to the teacher's teaching behavior, and trigger the feedback behavior pattern of the virtual child role. The AR visualization prompt submodule (1014) is used to display AR visualization prompts containing specific improvement measures in real time in virtual teaching scenarios.

4. The preschool teacher training supervision system according to claim 1, characterized in that, The multimodal sensing array module (102) includes: Multiple high-definition cameras are used to capture teachers' facial expressions and body movements; A depth sensor that provides three-dimensional motion trajectory data; A microphone array for collecting voice signals; and wearable biosensors for monitoring teachers' physiological indicators.

5. The preschool teacher training supervision system according to claim 1, characterized in that, The teaching behavior evaluation model is implemented using a multi-layer neural network architecture, including: Input layer: Used to receive teacher teaching characteristics and virtual child feedback data; Feature extraction layer: Contains multiple parallel convolutional neural network subnetworks, each processing teaching features of different modalities; Temporal modeling layer: Employs a long short-term memory temporal network to capture the dynamic temporal characteristics of teaching behavior; Decision fusion layer: Weighted fusion of multimodal temporal features through an attention mechanism; Evaluation output layer: A softmax classifier is used to map the fused features into a multi-dimensional evaluation result of teaching ability. The evaluation dimensions include the appropriateness of teaching strategies, the quality of teacher-student interaction, emotion management ability, and teaching innovation.

6. The preschool teacher training supervision system according to claim 1, characterized in that, The intelligent intervention module (104) includes: The strategy library (1041) stores improvement measures for different teaching problems; The improvement measures determination submodule (1042) identifies improvement measures corresponding to the teaching problem from the strategy library (1041) when the teacher's teaching ability assessment result indicates the existence of a teaching problem. The AR rendering submodule (1043) transforms the improvement measures into visual prompts in the virtual teaching scene and transmits them to the AR scenario simulation module (101).

7. The preschool teacher training supervision system according to claim 1, characterized in that, When providing visual prompts, the AR scenario simulation module (101) dynamically adjusts the intensity of the prompts based on the teacher's improvement.

8. A method for supervising the practical training of preschool teachers, characterized in that, Includes the following steps: A virtual preschool teaching scenario with adaptive teaching context is constructed, wherein the virtual preschool teaching scenario can dynamically adjust the interactive feedback of the virtual preschool character according to the teacher's teaching behavior; The system collects teachers' multimodal teaching characteristics in real time, including facial expressions, body movements, voice features, and physiological features. Based on the multimodal teaching characteristics of teachers and real-time feedback data from virtual child roles, the teaching ability of teachers is evaluated through a pre-set teaching behavior assessment model. Based on the evaluation results of teachers' teaching abilities, AR visual prompts containing specific improvement measures are generated and displayed in real time in the virtual teaching scenario.

9. The method for supervising the practical training of preschool teachers according to claim 8, characterized in that, It also includes the following steps: storing data from the teacher training process and generating teacher competency development reports through data analysis.

10. The method for supervising the practical training of preschool teachers according to claim 8, characterized in that, The teaching behavior evaluation model is trained using an end-to-end supervised learning framework, including: We collected a dataset of teaching behaviors from outstanding historical teachers, and labeled facial expressions, body movements, voice features, and physiological signals to form a sample of teaching behavior features with time-series labels. The input multimodal teaching features and virtual preschool feedback data are normalized and time-series aligned. Spatial features of each modality are extracted by parallel CNN sub-networks, cross-modal temporal dependencies are captured by LSTM layers, and multimodal feature weights are dynamically fused using an attention mechanism. During training, the real labels of the teacher evaluation dimension are used as supervision signals. The network parameters are optimized by the cross-entropy loss function. The weights of the CNN-LSTM-attention module are iteratively adjusted by the backpropagation algorithm. An early stopping strategy is adopted to prevent overfitting, so that the model can automatically learn the mapping relationship of teaching effectiveness from the temporal correlation between teacher behavior and virtual children's feedback.

Images