Microteaching actual effect evaluation system based on artificial intelligence

Abstract

The invention relates to the technical field of education, in particular to an artificial intelligence-based micro-teaching actual effect evaluation system, which comprises a micro-teaching recording module, a multi-modal data fusion module, a micro-teaching behavior analysis module, an emotion recognition feedback module, an intelligent evaluation generation module and a personalized teaching suggestion module. Wherein the micro-grid teaching recording module is used for recording teaching video and audio data of teachers in the micro-grid teaching process; the multi-modal data fusion module is used for revealing deep-level connection among different data sources; the micro teaching behavior analysis module is used for identifying teaching interaction points and student learning behavior modes; the emotion recognition feedback module is used for generating an emotion analysis result; according to the invention, by integrating multi-modal data analysis, emotion recognition and intelligent evaluation, the evaluation real-time performance and accuracy of micro teaching are improved, and at the same time, through personalized teaching, suggestions are improved, and the teaching strategy and the learning experience of students are effectively optimized.

Description

Technical Field

[0001] The present invention relates to the field of educational technology, and in particular to an artificial intelligence-based microteaching effectiveness evaluation system. Background Art

[0002] In today's educational environment, traditional teaching effectiveness evaluation methods usually rely on standardized test scores, teachers' subjective evaluations and students' self-reports. Although these methods provide a certain basis for evaluation, they are often unable to capture students' emotional state, concentration and the interaction between teachers and students in real time during the learning process. In addition, these traditional methods rarely provide specific suggestions for improving teachers' teaching strategies. Especially in the micro-teaching environment, this problem of lack of real-time feedback and personalized teaching suggestions is particularly prominent. As an emerging teaching model, micro-teaching has a unique structure that requires more refined and dynamic evaluation methods to optimize the teaching process and improve students' learning efficiency.

[0003] In view of the shortcomings of the existing technology, the present invention aims to solve the technical problem of how to evaluate the teaching effect in real time and comprehensively in a micro-teaching environment. In particular, the existing technology lacks a system that can comprehensively analyze students' emotional reactions, behavioral performances and learning interactions, which limits the instant adjustment and optimization of teaching strategies. In addition, the existing evaluation methods rarely provide specific and personalized teaching improvement suggestions based on the multi-dimensional data of teaching effects.

[0004] Therefore, developing an AI-based system that can integrate multimodal data analysis, emotion recognition, and intelligent assessment generation is not only of great significance for improving the quality and effectiveness of microteaching, but is also an inevitable trend in the development of educational technology. Summary of the invention

[0005] Based on the above objectives, the present invention provides a microteaching effectiveness evaluation system based on artificial intelligence.

[0006] The microteaching effectiveness evaluation system based on artificial intelligence includes a microteaching recording module, a multimodal data fusion module, a microteaching behavior analysis module, an emotion recognition feedback module, an intelligent evaluation generation module, and a personalized teaching suggestion module; among which:

[0007] Micro-teaching recording module: used to record the teacher's lecture video and audio data, as well as the students' interaction and feedback during the micro-teaching process, and generate detailed teaching records, which include video data, audio data and text data;

[0008] Multimodal data fusion module: It fuses the multi-source data from the micro-teaching recording module and uses the tensor decomposition algorithm to decompose the video, audio and text data into tensor form, thereby revealing the deep connection between different data sources and generating a unified data representation form;

[0009] Microteaching behavior analysis module: Receives unified data generated by the multimodal data fusion module, uses natural language processing and computer vision technology to conduct in-depth analysis of the fused data, identifies teaching interaction points and student learning behavior patterns, and generates behavior analysis results;

[0010] Emotion recognition feedback module: receives the behavior analysis results from the micro-teaching behavior analysis module, and uses emotion computing technology to analyze the students' emotional state in micro-teaching, including concentration, participation, and emotional fluctuations, generates emotion analysis results, and provides emotion feedback data;

[0011] Intelligent evaluation generation module: Integrates the emotional feedback data provided by the emotion recognition feedback module and the behavior analysis results of the micro-teaching behavior analysis module to generate a comprehensive micro-teaching effect evaluation report, which includes multiple dimensions such as teaching quality, student participation and knowledge mastery;

[0012] Personalized teaching improvement suggestion module: Based on the evaluation report provided by the intelligent evaluation generation module, the module uses machine learning algorithms to generate personalized teaching improvement suggestions for teachers, assisting teachers in optimizing teaching results.

[0013] Furthermore, the micro-teaching recording module includes a video acquisition unit, an audio acquisition unit, an interaction capture unit, a text recording unit and a data integration unit; wherein:

[0014] Video acquisition unit: used to capture visual information during the teacher's teaching process. By using a high-definition camera to record the teacher's movements, expressions, and blackboard content, the video data can reflect the full picture of the teaching scene.

[0015] Audio collection unit: uses a directional microphone to collect audio data from teachers’ lectures, and also captures students’ questions and feedback in class to ensure the clarity of the audio data;

[0016] Interaction Capture Unit: used to record students' interactive behaviors in class, including hand raising, facial expression changes and other non-verbal communication behaviors, and identify and record students' participation and interaction frequency through image recognition and sound analysis technology;

[0017] Text recording unit: used to transcribe the teacher's lectures and student feedback into text format, using speech recognition technology to automatically convert the teacher's and students' speeches into text records. At the same time, manually entered class notes will also be integrated into the record;

[0018] Data integration unit: Integrate the data collected by the video acquisition unit, audio acquisition unit, interaction capture unit and text recording unit to form a unified teaching record archive.

[0019] Furthermore, the multimodal data fusion module includes a preprocessing unit, a tensor construction unit, a tensor decomposition unit, a feature extraction unit and a data representation unit; wherein:

[0020] Preprocessing unit: preprocesses the video, audio and text data from the micro-teaching recording module, removes noise and standardizes the data format to ensure the quality and consistency of the data. Specifically, it adjusts the frame rate and enhances the image of the video data, performs noise reduction and spectrum analysis on the audio data, and performs word segmentation and semantic analysis on the text data.

[0021] Tensor construction unit: used to convert the preprocessed video, audio and text data into tensor form, specifically representing video data as a three-dimensional tensor, including time, space and color channels; representing audio data as a two-dimensional tensor, including time and frequency; and representing text data as a two-dimensional tensor, including word vectors and time series;

[0022] Tensor decomposition unit: used to decompose the constructed multi-modal tensor. Specifically, it uses the high-order singular value decomposition method to decompose the original high-dimensional tensor into multiple low-dimensional tensors and core tensors to reveal the deep connection between different data sources;

[0023] Feature extraction unit: used to extract features from the decomposed low-dimensional tensor and identify the correlation patterns between video, audio and text data. Specifically, feature screening and clustering algorithms are used to ensure that the extracted features are representative and relevant.

[0024] Data representation unit: used to integrate the extracted features into a unified data representation and generate a multimodal fusion feature vector.

[0025] Furthermore, the tensor decomposition unit includes:

[0026] High-dimensional tensor construction: The preprocessed video, audio, and text data are represented as three-dimensional tensors, two-dimensional tensors, and two-dimensional tensors, respectively. Specifically, the video tensor is The audio tensor is The text tensor is Among them, I is the number of frames of the video tensor; J is the frame height of the video tensor; K is the frame width of the video tensor; L is the number of time steps of the audio tensor; M is the number of frequency components of the audio tensor; N is the number of sentences of the text tensor; P is the word vector dimension of the text tensor;

[0027] Tensor expansion: Expand high-dimensional tensors into matrix form to facilitate subsequent decomposition processing. Specifically for video tensors Expand mode 1, mode 2 and mode 3 respectively to get the matrix V 1 ∈R I×(J·K) , V 2 ∈R J×(I·K) and V 3 ∈R K×(1·J) , the audio tensor and text tensor are also expanded;

[0028] Singular value decomposition: Perform singular value decomposition on each expanded matrix and expand the matrix V in mode 1 1 For example, execute Among them U 1 is a left singular matrix, S 1 is a diagonal matrix of singular values, W 1 is a right singular matrix;

[0029] High-order singular value decomposition: Integrate the above singular value decomposition results to obtain the core tensor of the video tensor: and the factor matrix The same approach is applied to the audio tensor and the text tensor;

[0030] Tensor reconstruction: Reconstruct the low-dimensional factor matrix and core tensor back to the original high-dimensional tensor form to ensure the information integrity during the tensor decomposition process. The reconstruction formula is: Where× n represents n-modular product;

[0031] Deep connection discovery: Through the process of tensor decomposition and reconstruction, we can identify and extract the potential patterns and association information between video, audio and text data. Represents the comprehensive characteristics of different data sources, the low-dimensional factor matrix U 1 , U 2 and U 3 It reflects the pattern of each data source, thereby revealing the deep connections between multimodal data sources.

[0032] Furthermore, the feature extraction unit includes:

[0033] Feature standardization: First, the data in the low-dimensional tensor from the high-order singular value decomposition is standardized to ensure the consistency of the data at different scales. Let T d It is a low-dimensional tensor, where d represents the data dimension, and the Zscore normalization formula is used: Among them, μ d It is T d The mean value, σ d is the standard deviation;

[0034] Feature screening: Apply the feature selection method based on information gain to select the most informative features for distinguishing different categories from the standardized data, calculate the information gain IG of each feature, and select the features with high information gain for analysis. The formula is: IG(T,F)=H(F)-H(T|F), where T is the target category, F is the feature, H(T) is the entropy of the target, and H(T|F) is the conditional entropy of the target after the feature is given;

[0035] Application of clustering algorithm: Use the k-means clustering algorithm to cluster the filtered features to identify the potential relationship between different patterns. The goal of the clustering algorithm is to minimize the sum of the squares of the distances from each cluster point to the cluster center. The formula is: Where k is the number of clusters, S i is the set of points in the ith cluster, μ i YesS i the center of;

[0036] Feature synthesis: The center point of each class in the clustering result is used as a new feature vector. The new feature vector represents the information of the original data and is used to reflect the association pattern between video, audio and text data.

[0037] Furthermore, the micro-teaching behavior analysis module includes a text analysis unit, a video analysis unit, an audio analysis unit and a behavior pattern recognition unit; wherein:

[0038] Text analysis unit: receives text data from the unified data generated by the multimodal data fusion module and analyzes it using natural language processing technology. Specifically, it processes text data using word segmentation, part-of-speech tagging, and named entity recognition methods, and uses a word vector model to convert text into vector representation, identifying keywords and key sentences in the teacher's lecture content and student feedback;

[0039] Video analysis unit: Receives video data from the unified data and analyzes it using computer vision technology. Specifically, it uses a convolutional neural network to extract features from video frames, recognizes the teacher's movements and expressions, and the students' non-verbal behaviors. Each frame of the image is processed by the convolution layer, pooling layer, and fully connected layer to obtain a feature vector.

[0040] Audio analysis unit: receives audio data from unified data, analyzes it using audio processing technology, converts audio signals into spectrograms through short-time Fourier transform, and processes the spectrograms using deep neural networks to extract audio features, including intonation, speech speed, and volume changes;

[0041] Behavior pattern recognition unit: It integrates the features extracted by the text analysis unit, video analysis unit and audio analysis unit, identifies the interaction points in teaching and students' learning behavior patterns through multimodal fusion technology, uses the preset multimodal deep learning model to perform feature fusion and behavior pattern recognition, and generates behavior analysis results.

[0042] Furthermore, the emotion recognition feedback module includes a concentration analysis unit, a participation analysis unit, an emotion fluctuation analysis unit and an emotion feedback generation unit; wherein:

[0043] Concentration analysis unit: receives the video features and audio features in the behavior analysis results, and uses facial expression recognition and eye tracking technology to analyze the student's concentration. Facial expression recognition detects changes in students' facial features through a convolutional neural network model, and eye tracking evaluates students' concentration by calculating changes in their gaze points. Specifically, the video feature is V f , the audio feature is A f , then the concentration analysis formula is:

[0044] Concentration = α CNN (V f )+β·EyeTracking(V f ), where α and β are weight coefficients, CNN(V f ) represents the facial expression recognition result, EyeTracking(V f ) represents the eye tracking results;

[0045] Participation analysis unit: receives audio features and text features from behavior analysis results, and uses speech sentiment analysis and text sentiment analysis techniques to evaluate students' participation. Speech sentiment analysis processes audio features through long short-term memory networks, and text sentiment analysis processes text features through BERT models to generate participation analysis results. Let the audio feature be A p , the text feature is T p , then the participation analysis formula is:

[0046] Participation = γ LSTM (A p )+δ·BERT(T p ), where γ and δ are weight coefficients, LSTM(A p ) represents the result of speech sentiment analysis, BERT(T p ) represents the result of text sentiment analysis;

[0047] Emotional fluctuation analysis unit: receives video features, audio features and text features from the behavior analysis results, comprehensively uses multimodal sentiment analysis technology to evaluate students' emotional fluctuations, and generates emotional fluctuation analysis results by fusing multi-source features through a multimodal deep learning model. Let the multimodal feature be M e=[V e ,A e ,T e ], the emotional fluctuation analysis formula is: emotional fluctuation = Transformer (M e );

[0048] Emotional feedback generation unit: Integrates the results of the concentration analysis unit, engagement analysis unit and emotional fluctuation analysis unit to generate a comprehensive emotional analysis report. The emotional analysis report includes detailed data and evaluation results of students' concentration, engagement and emotional fluctuations.

[0049] Furthermore, the intelligent assessment generation module includes a data integration unit, a multi-dimensional assessment unit, a report generation unit and a visualization unit; wherein:

[0050] Data integration unit: Receive and integrate the emotional feedback data from the emotion recognition feedback module and the behavior analysis results from the micro-teaching behavior analysis module. Specifically, the emotional data and behavior data are processed uniformly using data fusion technology to generate a complete evaluation data set. The emotional feedback data includes students' concentration, participation and emotional fluctuations, and the behavior analysis results include teaching interaction points and students' learning behavior patterns.

[0051] Multidimensional evaluation unit: conducts multidimensional evaluation on the integrated data set, including teaching quality evaluation, student engagement evaluation and knowledge mastery evaluation. Specifically, it uses the preset multidimensional analysis model to comprehensively analyze the emotional and behavioral data to generate evaluation indicators and scores for each dimension.

[0052] Report generation unit: Integrate the evaluation indicators and scores generated by the multi-dimensional evaluation unit to generate a comprehensive microteaching effect evaluation report. The report includes detailed evaluation results and total scores of each dimension, and provides specific improvement suggestions;

[0053] Visualization Unit: Visualize the data in the evaluation report, including charts and graphs, so that teachers and education administrators can intuitively understand the evaluation results.

[0054] Furthermore, the multidimensional evaluation unit includes:

[0055] Data standardization: First, the integrated emotional data and behavioral data are standardized to ensure that the data of different dimensions have a consistent scale. Let the emotional data be E and the behavioral data be B. The standardization formula is:

[0056] and Among them, μ E and σ E are the mean and standard deviation of sentiment data, μ B and σ Bare the mean and standard deviation of the behavioral data, respectively;

[0057] Feature extraction: Use principal component analysis to extract features from the standardized data to reduce the data dimension and retain the main information. The calculation formula is: Z = XW, where X is the standardized data matrix, W is the weight matrix containing the principal component direction, and Z is the feature matrix after dimensionality reduction;

[0058] Emotional evaluation: Based on the extracted emotional features, a multidimensional sentiment analysis model is used to evaluate the students’ emotional state. The emotional evaluation formula is:

[0059] Emotion score = α·AttentionScore+β·EngagementScore+γ·EmotionScore; where α, β and γ are weight coefficients, corresponding to the scores of concentration, engagement and emotional fluctuations respectively;

[0060] Behavior assessment: Based on the extracted behavioral characteristics, a multidimensional behavior analysis model is used to evaluate students’ learning behavior. The behavior assessment formula is: Among them, w i is the weight of the i-th behavior feature, BehaviorFeature i is the value of the i-th behavior feature;

[0061] Comprehensive evaluation: The emotional score and behavioral score are comprehensively analyzed to generate evaluation indicators and total scores for each dimension. The comprehensive evaluation formula is: comprehensive score = λ·emotional score + δ·behavioral score, where λ and δ are the weight coefficients of the emotional score and behavioral score.

[0062] Furthermore, the personalized teaching improvement suggestion module includes a data receiving unit, a feature extraction unit, a model training unit, a strategy generation unit and a resource recommendation unit; wherein:

[0063] Data receiving unit: receiving the comprehensive evaluation report provided by the intelligent evaluation generation module, which includes evaluation indicators and scores in multiple dimensions of teaching quality, student participation and knowledge mastery;

[0064] Feature extraction unit: extracts features from the evaluation report, including the teacher's teaching behavior, students' emotional feedback and learning behavior patterns, to generate personalized teaching improvement suggestions;

[0065] Model training unit: Using historical teaching data and corresponding teaching effect evaluation, the machine learning model is trained through supervised learning methods. The model training unit includes three steps: data preprocessing, feature selection, and model training, and finally generates a model that can predict the best teaching strategies and resources;

[0066] Strategy generation unit: Based on the trained machine learning model, it analyzes the characteristics of the current teaching evaluation report and generates specific teaching strategy suggestions, including teaching method adjustment, classroom interaction optimization and time management strategy;

[0067] Resource recommendation unit: Use collaborative filtering algorithms to recommend appropriate teaching resources based on the needs of teachers and students. Teaching resources include teaching videos, courseware, exercise banks, and reference materials.

[0068] Beneficial effects of the present invention:

[0069] The present invention integrates multimodal data fusion technology, emotion recognition, and behavior analysis technology, so that teachers can instantly understand students' emotional state, concentration, and participation in class. This comprehensive evaluation capability not only helps teachers adjust teaching methods in a timely manner to better meet students' learning needs, but also provides personalized feedback based on each student's specific performance, thereby optimizing learning outcomes.

[0070] The present invention, through the designed intelligent evaluation generation module, can generate detailed teaching effect reports according to comprehensive evaluation data, including evaluations of teaching quality, student participation and knowledge mastery. Based on these reports, the personalized teaching improvement suggestion module further provides specific teaching strategies and resource recommendations to support teachers in making targeted improvements in teaching practice. This systematic optimization not only improves teaching efficiency, but also enhances the implementation effect of teaching strategies, which helps to improve the overall education quality and student satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0072] Figure 1 A schematic diagram of a microteaching effectiveness evaluation system according to an embodiment of the present invention;

[0073] Figure 2 Schematic diagram of an intelligent assessment generation module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0075] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0076] like Figure 1-Figure 2 As shown in the figure, the micro-teaching effectiveness evaluation system based on artificial intelligence includes a micro-teaching recording module, a multimodal data fusion module, a micro-teaching behavior analysis module, an emotion recognition feedback module, an intelligent evaluation generation module and a personalized teaching suggestion module; among which:

[0077] Micro-teaching recording module: used to record the teacher's lecture video and audio data, as well as the students' interaction and feedback during micro-teaching, and generate detailed teaching records, which include video data, audio data and text data;

[0078] Multimodal data fusion module: It fuses the multi-source data from the micro-teaching recording module and uses the tensor decomposition algorithm to decompose the video, audio and text data into tensor form, thereby revealing the deep connection between different data sources and generating a unified data representation form;

[0079] Microteaching behavior analysis module: Receives unified data generated by the multimodal data fusion module, uses natural language processing and computer vision technology to conduct in-depth analysis of the fused data, identifies teaching interaction points and student learning behavior patterns, and generates behavior analysis results;

[0080] Emotion recognition feedback module: receives the behavior analysis results from the micro-teaching behavior analysis module, and uses emotion computing technology to analyze the students' emotional state in micro-teaching, including concentration, participation, and emotional fluctuations, generates emotion analysis results, and provides emotion feedback data;

[0081] Intelligent evaluation generation module: Integrates the emotional feedback data provided by the emotion recognition feedback module and the behavior analysis results of the micro-teaching behavior analysis module to generate a comprehensive micro-teaching effect evaluation report, which includes multiple dimensions such as teaching quality, student participation and knowledge mastery;

[0082] Personalized teaching improvement suggestion module: Based on the evaluation report provided by the intelligent evaluation generation module, the module uses machine learning algorithms to generate personalized teaching improvement suggestions for teachers, assisting teachers in optimizing teaching results.

[0083] The micro-teaching recording module includes a video acquisition unit, an audio acquisition unit, an interaction capture unit, a text recording unit, and a data integration unit; among which:

[0084] Video acquisition unit: used to capture visual information during the teacher's teaching process. By using a high-definition camera to record the teacher's movements, expressions, and blackboard content, the video data can reflect the full picture of the teaching scene.

[0085] Audio collection unit: uses a directional microphone to collect audio data from teachers’ lectures, and also captures students’ questions and feedback in class to ensure the clarity of the audio data;

[0086] Interaction Capture Unit: used to record students' interactive behaviors in class, including hand raising, facial expression changes and other non-verbal communication behaviors, and identify and record students' participation and interaction frequency through image recognition and sound analysis technology;

[0087] Text recording unit: used to transcribe the teacher's lectures and student feedback into text format, using speech recognition technology to automatically convert the teacher's and students' speeches into text records. At the same time, manually entered class notes will also be integrated into the record;

[0088] Data integration unit: Integrate the data collected by the video acquisition unit, audio acquisition unit, interaction capture unit and text recording unit to form a unified teaching record archive. This data integration process ensures that the data format of all records is unified, which is convenient for subsequent processing and analysis. Through this above-mentioned recording method, the micro-teaching recording module can provide comprehensive and in-depth teaching data, support more accurate teaching analysis and evaluation, and integrate multiple types of data. Not only does it enhance the accuracy of teaching evaluation, it also enables teachers to obtain detailed feedback on teaching interactions and student participation, further optimizing teaching strategies and improving teaching quality.

[0089] The multimodal data fusion module includes a preprocessing unit, a tensor construction unit, a tensor decomposition unit, a feature extraction unit and a data representation unit; wherein:

[0090] Preprocessing unit: preprocesses the video, audio and text data from the micro-teaching recording module, removes noise and standardizes the data format to ensure the quality and consistency of the data. Specifically, it adjusts the frame rate and enhances the image of the video data, performs noise reduction and spectrum analysis on the audio data, and performs word segmentation and semantic analysis on the text data.

[0091] Tensor construction unit: used to convert the preprocessed video, audio and text data into tensor form, specifically representing video data as a three-dimensional tensor, including time, space and color channels; representing audio data as a two-dimensional tensor, including time and frequency; and representing text data as a two-dimensional tensor, including word vectors and time series;

[0092] Tensor decomposition unit: used to decompose the constructed multimodal tensor. Specifically, the high-order singular value decomposition (HOSVD) method is used to decompose the original high-dimensional tensor into multiple low-dimensional tensors and core tensors to reveal the deep connection between different data sources.

[0093] Feature extraction unit: used to extract features from the decomposed low-dimensional tensor and identify the correlation patterns between video, audio and text data. Specifically, feature screening and clustering algorithms are used to ensure that the extracted features are representative and relevant.

[0094] Data representation unit: used to integrate the extracted features into a unified data representation and generate a multimodal fusion feature vector, which not only retains the unique information of each data source, but also reflects the correlation between them, providing a basis for subsequent analysis and evaluation; through the description of the above multimodal data fusion module, this module can effectively fuse data from different sources, use tensor decomposition algorithm to deeply reveal the complex relationship between video, audio and text data, and generate a unified and meaningful data representation. This processing method not only improves the accuracy and efficiency of data fusion, but also provides a solid data foundation for teaching behavior analysis and emotion recognition, thereby improving the overall evaluation effect of the system.

[0095] The tensor decomposition unit includes:

[0096] High-dimensional tensor construction: The preprocessed video, audio, and text data are represented as three-dimensional tensors, two-dimensional tensors, and two-dimensional tensors, respectively. Specifically, the video tensor is The audio tensor is The text tensor is Among them, I is the number of frames of the video tensor; J is the frame height of the video tensor; K is the frame width of the video tensor; L is the number of time steps of the audio tensor; M is the number of frequency components of the audio tensor; N is the number of sentences of the text tensor; P is the word vector dimension of the text tensor;

[0097] Tensor expansion: Expand high-dimensional tensors into matrix form to facilitate subsequent decomposition processing. Specifically for video tensors Expand mode 1, mode 2 and mode 3 respectively to get the matrix V 1 ∈R I×(J·K) , V 2 ∈R J×(I·K) and V 3 ∈RK×(1·J) , the audio tensor and text tensor are also expanded;

[0098] Singular value decomposition: Perform singular value decomposition on each expanded matrix and expand the matrix V in mode 1 1 For example, execute Among them U 1 is a left singular matrix, S 1 is a diagonal matrix of singular values, W 1 is a right singular matrix;

[0099] High-order singular value decomposition: Integrate the above singular value decomposition results to obtain the core tensor of the video tensor: and the factor matrix The same approach is applied to the audio tensor and the text tensor;

[0100] Tensor reconstruction: Reconstruct the low-dimensional factor matrix and core tensor back to the original high-dimensional tensor form to ensure the information integrity during the tensor decomposition process. The reconstruction formula is: Where× n represents n-modular product;

[0101] Deep connection discovery: Through the process of tensor decomposition and reconstruction, we can identify and extract the potential patterns and association information between video, audio and text data. Represents the comprehensive characteristics of different data sources, the low-dimensional factor matrix U 1 , U 2 and U 3 It reflects the pattern of each data source, thereby revealing the deep connection between multimodal data sources; through the above detailed tensor decomposition process, this module can effectively decompose multi-source data into core tensors and low-dimensional factor matrices, and deeply reveal the complex relationship between different data sources. This processing method not only improves the efficiency and accuracy of data fusion, but also provides a richer and more accurate data basis for subsequent behavior analysis and emotion recognition, thereby improving the evaluation effect and reliability of the entire system.

[0102] The feature extraction unit includes:

[0103] Feature standardization: First, the data in the low-dimensional tensor from the high-order singular value decomposition is standardized to ensure the consistency of the data at different scales. Let T d It is a low-dimensional tensor, where d represents the data dimension, and the Zscore normalization formula is used: Among them, μ d It is T d The mean value, σ d is the standard deviation;

[0104] Feature screening: Apply the feature selection method based on information gain to select the most informative features for distinguishing different categories from the standardized data, calculate the information gain IG of each feature, and select the features with high information gain for analysis. The formula is: IG(T,F)=H(T)-H(T|F), where T is the target category, F is the feature, H(T) is the entropy of the target, and H(T|F) is the conditional entropy of the target after the feature is given;

[0105] Application of clustering algorithm: Use the k-means clustering algorithm to cluster the filtered features to identify the potential relationship between different patterns. The goal of the clustering algorithm is to minimize the sum of the squares of the distances from each cluster point to the cluster center. The formula is: Where k is the number of clusters, S i is the set of points in the ith cluster, μ i YesS i the center of;

[0106] Feature synthesis: The center point of each class in the clustering result is used as a new feature vector. The new feature vector represents the information of the original data and is used to reflect the correlation pattern between video, audio and text data. Through the above feature extraction process, the feature extraction unit can effectively extract key features from low-dimensional tensors and reveal the deep connection between data through clustering analysis. This not only enhances the depth and breadth of data analysis, but also provides the system with more accurate data interpretation and more effective decision support. These efficient feature extraction and data processing strategies greatly improve the system's operating efficiency and analysis accuracy.

[0107] The microteaching behavior analysis module includes a text analysis unit, a video analysis unit, an audio analysis unit, and a behavior pattern recognition unit; among which:

[0108] Text analysis unit: receives text data from the unified data generated by the multimodal data fusion module and analyzes it using natural language processing technology. Specifically, it processes text data using word segmentation, part-of-speech tagging, and named entity recognition methods, and uses a word vector model (such as Word2Vec) to convert text into vector representation, identifying keywords and key sentences in the teacher's lecture content and student feedback;

[0109] Video analysis unit: Receives video data from the unified data and analyzes it using computer vision technology. Specifically, it uses a convolutional neural network (CNN) to extract features from video frames, recognizes the teacher's movements and expressions, and the students' non-verbal behaviors (such as hand raising and nodding). Each frame of the image is processed by the convolution layer, pooling layer, and fully connected layer to obtain a feature vector.

[0110] Audio analysis unit: receives audio data from the unified data, analyzes it using audio processing technology, converts the audio signal into a spectrogram through short-time Fourier transform (STFT), and processes the spectrogram using a deep neural network to extract audio features, including intonation, speech speed, and volume changes;

[0111] Behavior pattern recognition unit: integrates the features extracted by the text analysis unit, video analysis unit and audio analysis unit, identifies the interaction points in teaching and the students' learning behavior patterns through multimodal fusion technology, uses the preset multimodal deep learning model (such as multimodal LSTM) to perform feature fusion and behavior pattern recognition, and generates behavior analysis results; through the detailed description of the above micro-teaching behavior analysis module, this module can comprehensively receive and process multimodal data, use natural language processing and computer vision technology to deeply analyze the text, video and audio data in the teaching process, through feature extraction and multimodal fusion, this module can accurately identify the interaction points in teaching and the students' learning behavior patterns, and generate detailed behavior analysis results, thereby providing high-quality data support for subsequent emotion recognition and feedback and intelligent evaluation, and improving the analysis and evaluation effect of the entire system.

[0112] The emotion recognition feedback module includes a concentration analysis unit, an engagement analysis unit, an emotion fluctuation analysis unit, and an emotion feedback generation unit; wherein:

[0113] Concentration analysis unit: receives the video features and audio features in the behavior analysis results, and uses facial expression recognition and eye tracking technology to analyze the student's concentration. Facial expression recognition detects changes in students' facial features through a convolutional neural network model, and eye tracking evaluates students' concentration by calculating changes in their gaze points. Specifically, the video feature is V f , the audio feature is A f , then the concentration analysis formula is:

[0114] Concentration = α CNN (V f )+β·EyeTracking(V f ), where α and β are weight coefficients, CNN(V f ) represents the facial expression recognition result, EyeTracking(V f ) represents the eye tracking results;

[0115] Participation analysis unit: receives audio features and text features from behavior analysis results, and uses speech sentiment analysis and text sentiment analysis techniques to evaluate students' participation. Speech sentiment analysis processes audio features through long short-term memory networks, and text sentiment analysis processes text features through BERT models to generate participation analysis results. Let the audio feature be A p , the text feature is T p, then the participation analysis formula is:

[0116] Participation = γ LSTM (A p )+δ·BERT(T p ), where γ and δ are weight coefficients, LSTM(A p ) represents the result of speech sentiment analysis, BERT(T p ) represents the result of text sentiment analysis;

[0117] Emotional fluctuation analysis unit: receives video features, audio features and text features from the behavior analysis results, comprehensively uses multimodal sentiment analysis technology to evaluate students' emotional fluctuations, and generates emotional fluctuation analysis results by fusing multi-source features through a multimodal deep learning model. Let the multimodal feature be M e =[V e ,A e ,T e ], the emotional fluctuation analysis formula is: emotional fluctuation = Transformer (M e );

[0118] Emotional feedback generation unit: Integrate the results of the concentration analysis unit, the engagement analysis unit and the emotional fluctuation analysis unit to generate a comprehensive emotional analysis report. The emotional analysis report includes detailed data and evaluation results of students' concentration, engagement and emotional fluctuations. Through the detailed description of the above-mentioned emotion recognition feedback module, this module can fully receive and process the behavior analysis results, use advanced emotion computing technology to conduct in-depth analysis of students' emotional state in micro-teaching, and generate detailed emotional analysis reports through multi-dimensional evaluation of concentration, engagement and emotional fluctuations, so as to help teachers understand students' emotional state in real time and adjust teaching strategies, thereby optimizing teaching effects and enhancing students' learning experience.

[0119] The intelligent assessment generation module includes a data integration unit, a multi-dimensional assessment unit, a report generation unit and a visualization unit; wherein:

[0120] Data integration unit: Receive and integrate the emotional feedback data from the emotion recognition feedback module and the behavior analysis results from the micro-teaching behavior analysis module. Specifically, the emotional data and behavior data are processed uniformly using data fusion technology to generate a complete evaluation data set. The emotional feedback data includes students' concentration, participation and emotional fluctuations, and the behavior analysis results include teaching interaction points and students' learning behavior patterns.

[0121] Multidimensional evaluation unit: conducts multidimensional evaluation on the integrated data set, including teaching quality evaluation, student engagement evaluation and knowledge mastery evaluation. Specifically, it uses the preset multidimensional analysis model to comprehensively analyze the emotional and behavioral data to generate evaluation indicators and scores for each dimension.

[0122] Report generation unit: Integrate the evaluation indicators and scores generated by the multi-dimensional evaluation unit to generate a comprehensive microteaching effect evaluation report. The report includes detailed evaluation results and total scores of each dimension, and provides specific improvement suggestions. This unit uses template generation technology to ensure the standardization and readability of the evaluation report.

[0123] Visualization unit: Visualize the data in the evaluation report, including charts and graphs, so that teachers and educational administrators can intuitively understand the evaluation results. Specifically, data visualization technology is used to convert complex data into concise and clear graphical representations. Through the detailed description of the above-mentioned intelligent evaluation generation module, this module can comprehensively integrate the data provided by the emotion recognition feedback module and the micro-teaching behavior analysis module, and use multi-dimensional evaluation technology to generate a comprehensive micro-teaching effectiveness evaluation report. The evaluation report not only provides detailed evaluation results of teaching quality, student participation and knowledge mastery, but also makes the evaluation results more intuitive and easy to understand through data visualization technology, which helps teachers and educational administrators to adjust teaching strategies in a timely manner, optimize teaching effects and improve teaching quality.

[0124] The multi-dimensional assessment unit includes:

[0125] Data standardization: First, the integrated emotional data and behavioral data are standardized to ensure that the data of different dimensions have consistent scales. Let the emotional data be E (including concentration, participation, and emotional fluctuations) and the behavioral data be B (including teaching interaction points and learning behavior patterns). The standardization formula is:

[0126] and Among them, μ E and σ E are the mean and standard deviation of sentiment data, μ B and σ B are the mean and standard deviation of the behavioral data, respectively;

[0127] Feature extraction: Use principal component analysis to extract features from the standardized data to reduce the data dimension and retain the main information. The calculation formula is: Z = XW, where X is the standardized data matrix, W is the weight matrix containing the principal component direction, and Z is the feature matrix after dimensionality reduction;

[0128] Emotional evaluation: Based on the extracted emotional features, a multidimensional sentiment analysis model is used to evaluate the students’ emotional state. The emotional evaluation formula is:

[0129] Emotion score = α·AttentionScore+β·EngagementScore+γ·EmotionScore; where α, β and γ are weight coefficients, corresponding to the scores of concentration, engagement and emotional fluctuations respectively;

[0130] Behavior assessment: Based on the extracted behavioral characteristics, a multidimensional behavior analysis model is used to evaluate students’ learning behavior. The behavior assessment formula is: Among them, w i is the weight of the i-th behavior feature, BehaviorFeature i is the value of the i-th behavior feature;

[0131] Comprehensive evaluation: The emotional score and behavioral score are comprehensively analyzed to generate evaluation indicators and total scores for each dimension. The comprehensive evaluation formula is: Comprehensive score = λ·emotional score + δ·behavioral score, where λ and δ are weight coefficients of the emotional score and behavioral score. Through the detailed description of the above multidimensional evaluation unit, the unit can use the preset multidimensional analysis model to standardize, extract features and conduct multidimensional evaluation of the integrated emotional data and behavioral data to generate detailed evaluation indicators and scores. Through the comprehensive analysis of the emotional score and behavioral score, the evaluation report can comprehensively reflect the teaching quality, student participation and knowledge mastery, and provide valuable feedback to teachers and education administrators to help them optimize teaching strategies and improve teaching effectiveness.

[0132] The personalized teaching improvement suggestion module includes a data receiving unit, a feature extraction unit, a model training unit, a strategy generation unit, and a resource recommendation unit; among which:

[0133] Data receiving unit: receiving the comprehensive evaluation report provided by the intelligent evaluation generation module, which includes evaluation indicators and scores in multiple dimensions of teaching quality, student participation and knowledge mastery;

[0134] Feature extraction unit: extracts features from the evaluation report, including the teacher's teaching behavior, students' emotional feedback and learning behavior patterns, to generate personalized teaching improvement suggestions;

[0135] Model training unit: Using historical teaching data and corresponding teaching effect evaluation, the machine learning model is trained through supervised learning methods. The model training unit includes three steps: data preprocessing, feature selection, and model training, and finally generates a model that can predict the best teaching strategies and resources;

[0136] Strategy generation unit: Based on the trained machine learning model, it analyzes the characteristics of the current teaching evaluation report and generates specific teaching strategy suggestions, including teaching method adjustment, classroom interaction optimization and time management strategy;

[0137] Resource recommendation unit: using collaborative filtering algorithms, recommending appropriate teaching resources based on the needs of teachers and students. Teaching resources include teaching videos, courseware, exercise banks and reference materials, aiming to help teachers improve teaching effectiveness and enhance students' learning experience. Through the detailed description of the personalized teaching improvement suggestion module mentioned above, this module can use the comprehensive evaluation report provided by the intelligent evaluation generation module to generate personalized teaching improvement suggestions for teachers through machine learning algorithms. The improvement suggestions include specific teaching strategies and resource recommendations to help teachers optimize teaching effectiveness and improve teaching quality. Through the collaborative work of five units including data reception, feature extraction, model training, strategy generation and resource recommendation, the system can provide comprehensive and accurate teaching improvement suggestions to enhance teachers' teaching ability and students' learning effects.

[0138] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. The microteaching effectiveness evaluation system based on artificial intelligence is characterized by: It includes micro-teaching recording module, multimodal data fusion module, micro-teaching behavior analysis module, emotion recognition feedback module, intelligent evaluation generation module and personalized teaching suggestion module; among which: Micro-teaching recording module: used to record the teacher's lecture video and audio data, as well as the students' interaction and feedback during the micro-teaching process, and generate detailed teaching records, which include video data, audio data and text data; Multimodal data fusion module: It fuses the multi-source data from the micro-teaching recording module and uses the tensor decomposition algorithm to decompose the video, audio and text data into tensor form, thereby revealing the deep connection between different data sources and generating a unified data representation form; Microteaching behavior analysis module: Receives unified data generated by the multimodal data fusion module, uses natural language processing and computer vision technology to conduct in-depth analysis of the fused data, identifies teaching interaction points and student learning behavior patterns, and generates behavior analysis results; Emotion recognition feedback module: receives the behavior analysis results from the micro-teaching behavior analysis module, and uses emotion computing technology to analyze the students' emotional state in micro-teaching, including concentration, participation, and emotional fluctuations, generates emotion analysis results, and provides emotion feedback data; Intelligent evaluation generation module: Integrates the emotional feedback data provided by the emotion recognition feedback module and the behavior analysis results of the micro-teaching behavior analysis module to generate a comprehensive micro-teaching effect evaluation report, which includes multiple dimensions such as teaching quality, student participation and knowledge mastery; Personalized teaching improvement suggestion module: Based on the evaluation report provided by the intelligent evaluation generation module, the module uses machine learning algorithms to generate personalized teaching improvement suggestions for teachers, assisting teachers in optimizing teaching results.

2. The microteaching effectiveness evaluation system based on artificial intelligence according to claim 1 is characterized in that: The micro-teaching recording module includes a video acquisition unit, an audio acquisition unit, an interaction capture unit, a text recording unit and a data integration unit; wherein: Video acquisition unit: used to capture visual information during the teacher's teaching process. By using a high-definition camera to record the teacher's movements, expressions, and blackboard content, the video data can reflect the full picture of the teaching scene. Audio collection unit: uses a directional microphone to collect audio data from teachers’ lectures, and also captures students’ questions and feedback in class to ensure the clarity of the audio data; Interaction Capture Unit: used to record students' interactive behaviors in class, including hand raising, facial expression changes and other non-verbal communication behaviors, and identify and record students' participation and interaction frequency through image recognition and sound analysis technology; Text recording unit: used to transcribe the teacher's lectures and student feedback into text format, using speech recognition technology to automatically convert the teacher's and students' speeches into text records. At the same time, manually entered class notes will also be integrated into the record; Data integration unit: Integrate the data collected by the video acquisition unit, audio acquisition unit, interaction capture unit and text recording unit to form a unified teaching record archive.

3. The microteaching effectiveness evaluation system based on artificial intelligence according to claim 1 is characterized in that: The multimodal data fusion module includes a preprocessing unit, a tensor construction unit, a tensor decomposition unit, a feature extraction unit and a data representation unit; wherein: Preprocessing unit: preprocesses the video, audio and text data from the micro-teaching recording module, removes noise and standardizes the data format to ensure the quality and consistency of the data. Specifically, it adjusts the frame rate and enhances the image of the video data, performs noise reduction and spectrum analysis on the audio data, and performs word segmentation and semantic analysis on the text data. Tensor construction unit: used to convert the preprocessed video, audio and text data into tensor form, specifically representing video data as a three-dimensional tensor, including time, space and color channels; representing audio data as a two-dimensional tensor, including time and frequency; and representing text data as a two-dimensional tensor, including word vectors and time series; Tensor decomposition unit: used to decompose the constructed multi-modal tensor. Specifically, it uses the high-order singular value decomposition method to decompose the original high-dimensional tensor into multiple low-dimensional tensors and core tensors to reveal the deep connection between different data sources; Feature extraction unit: used to extract features from the decomposed low-dimensional tensor and identify the correlation patterns between video, audio and text data. Specifically, feature screening and clustering algorithms are used to ensure that the extracted features are representative and relevant. Data representation unit: used to integrate the extracted features into a unified data representation and generate a multimodal fusion feature vector.

4. The microteaching effectiveness evaluation system based on artificial intelligence according to claim 3 is characterized in that: The tensor decomposition unit comprises: High-dimensional tensor construction: The preprocessed video, audio, and text data are represented as three-dimensional tensors, two-dimensional tensors, and two-dimensional tensors, respectively. Specifically, the video tensor is The audio tensor is The text tensor is Among them, I is the number of frames of the video tensor; J is the frame height of the video tensor; K is the frame width of the video tensor; L is the number of time steps of the audio tensor; M is the number of frequency components of the audio tensor; N is the number of sentences of the text tensor; P is the word vector dimension of the text tensor; Tensor expansion: Expand high-dimensional tensors into matrix form to facilitate subsequent decomposition processing. Specifically for video tensors Expand mode 1, mode 2 and mode 3 respectively to obtain the matrix V1∈R I×(J·K) , V2∈R J×(I·K) and V3∈R K×(1·J) , the audio tensor and text tensor are also expanded; Singular value decomposition: Perform singular value decomposition on each expanded matrix. Take the expanded matrix V1 of mode 1 as an example and execute Where U1 is the left singular matrix, S1 is the singular value diagonal matrix, and W1 is the right singular matrix; High-order singular value decomposition: Integrate the above singular value decomposition results to obtain the core tensor of the video tensor: and the factor matrix The same approach is applied to the audio tensor and the text tensor; Tensor reconstruction: Reconstruct the low-dimensional factor matrix and core tensor back to the original high-dimensional tensor form to ensure the information integrity during the tensor decomposition process. The reconstruction formula is: Where ×n represents n-module product; Deep connection discovery: Through the process of tensor decomposition and reconstruction, we can identify and extract the potential patterns and association information between video, audio and text data. Representing the comprehensive characteristics of different data sources, the low-dimensional factor matrices U1, U2, and U3 reflect the pattern of each data source, thereby revealing the deep connections between multimodal data sources.

5. The microteaching effectiveness evaluation system based on artificial intelligence according to claim 4 is characterized in that: The feature extraction unit comprises: Feature standardization: First, the data in the low-dimensional tensor from the high-order singular value decomposition is standardized to ensure the consistency of the data at different scales. Let T d It is a low-dimensional tensor, where d represents the data dimension, and the Zscore normalization formula is used: Among them, μ d It is T d The mean value, σ d is the standard deviation; Feature screening: Apply the feature selection method based on information gain to select the most informative features for distinguishing different categories from the standardized data, calculate the information gain IG of each feature, and select the features with high information gain for analysis. The formula is: IG(T,F)=H(T)-H(T|F), where T is the target category, F is the feature, H(T) is the entropy of the target, and H(T|F) is the conditional entropy of the target after the feature is given; Application of clustering algorithm: Use the k-means clustering algorithm to cluster the filtered features to identify the potential relationship between different patterns. The goal of the clustering algorithm is to minimize the sum of the squares of the distances from each cluster point to the cluster center. The formula is: Where k is the number of clusters, S i is the set of points in the ith cluster, μ i YesS i the center of; Feature synthesis: The center point of each class in the clustering result is used as a new feature vector. The new feature vector represents the information of the original data and is used to reflect the association pattern between video, audio and text data.

6. The microteaching effectiveness evaluation system based on artificial intelligence according to claim 1 is characterized in that: The micro-teaching behavior analysis module includes a text analysis unit, a video analysis unit, an audio analysis unit and a behavior pattern recognition unit; wherein: Text analysis unit: receives text data from the unified data generated by the multimodal data fusion module and analyzes it using natural language processing technology. Specifically, it processes text data using word segmentation, part-of-speech tagging, and named entity recognition methods, and uses a word vector model to convert text into vector representation, identifying keywords and key sentences in the teacher's lecture content and student feedback; Video analysis unit: Receives video data from the unified data and analyzes it using computer vision technology. Specifically, it uses a convolutional neural network to extract features from video frames, recognizes the teacher's movements and expressions, and the students' non-verbal behaviors. Each frame of the image is processed by the convolution layer, pooling layer, and fully connected layer to obtain a feature vector. Audio analysis unit: receives audio data from unified data, analyzes it using audio processing technology, converts audio signals into spectrograms through short-time Fourier transform, and processes the spectrograms using deep neural networks to extract audio features, including intonation, speech speed, and volume changes; Behavior pattern recognition unit: It integrates the features extracted by the text analysis unit, video analysis unit and audio analysis unit, identifies the interaction points in teaching and students' learning behavior patterns through multimodal fusion technology, uses the preset multimodal deep learning model to perform feature fusion and behavior pattern recognition, and generates behavior analysis results.

7. The microteaching effectiveness evaluation system based on artificial intelligence according to claim 1 is characterized in that: The emotion recognition feedback module includes a concentration analysis unit, a participation analysis unit, an emotion fluctuation analysis unit and an emotion feedback generation unit; wherein: Concentration analysis unit: receives the video features and audio features in the behavior analysis results, and uses facial expression recognition and eye tracking technology to analyze the student's concentration. Facial expression recognition detects changes in students' facial features through a convolutional neural network model, and eye tracking evaluates their concentration by calculating changes in the student's gaze point. Specifically, the video feature is V f , the audio feature is A f , then the concentration analysis formula is: Concentration = α CNN (V f )+β·EyeTracking(V f ), where α and β are weight coefficients, CNN(V f ) represents the facial expression recognition result, EyeTracking(V f ) represents the eye tracking results; Participation analysis unit: receives audio features and text features from behavior analysis results, and uses speech sentiment analysis and text sentiment analysis techniques to evaluate students' participation. Speech sentiment analysis processes audio features through long short-term memory networks, and text sentiment analysis processes text features through BERT models to generate participation analysis results. Let the audio feature be A p , the text feature is T p , then the participation analysis formula is: Participation = γ LSTM (A p )+δ·BERT(T p ), where γ and _ are weight coefficients, LSTM(A p ) represents the result of speech sentiment analysis, BERT(T p ) represents the result of text sentiment analysis; Emotional fluctuation analysis unit: receives video features, audio features and text features from the behavior analysis results, comprehensively uses multimodal sentiment analysis technology to evaluate students' emotional fluctuations, and generates emotional fluctuation analysis results by fusing multi-source features through a multimodal deep learning model. Let the multimodal feature be M e =[V e ,A e ,T e ], the emotional fluctuation analysis formula is: emotional fluctuation = Transformer (M e ); Emotional feedback generation unit: Integrates the results of the concentration analysis unit, engagement analysis unit and emotional fluctuation analysis unit to generate a comprehensive emotional analysis report. The emotional analysis report includes detailed data and evaluation results of students' concentration, engagement and emotional fluctuations.

8. The microteaching effectiveness evaluation system based on artificial intelligence according to claim 1 is characterized in that: The intelligent assessment generation module includes a data integration unit, a multi-dimensional assessment unit, a report generation unit and a visualization unit; wherein: Data integration unit: Receive and integrate the emotional feedback data from the emotion recognition feedback module and the behavior analysis results from the micro-teaching behavior analysis module. Specifically, the emotional data and behavior data are processed uniformly using data fusion technology to generate a complete evaluation data set. The emotional feedback data includes students' concentration, participation and emotional fluctuations, and the behavior analysis results include teaching interaction points and students' learning behavior patterns. Multidimensional evaluation unit: conducts multidimensional evaluation on the integrated data set, including teaching quality evaluation, student engagement evaluation and knowledge mastery evaluation. Specifically, it uses the preset multidimensional analysis model to comprehensively analyze the emotional and behavioral data to generate evaluation indicators and scores for each dimension. Report generation unit: Integrate the evaluation indicators and scores generated by the multi-dimensional evaluation unit to generate a comprehensive microteaching effect evaluation report. The report includes detailed evaluation results and total scores of each dimension, and provides specific improvement suggestions; Visualization Unit: Visualize the data in the evaluation report, including charts and graphs, so that teachers and education administrators can intuitively understand the evaluation results.

9. The microteaching effectiveness evaluation system based on artificial intelligence according to claim 8 is characterized in that: The multidimensional evaluation unit comprises: Data standardization: First, the integrated emotional data and behavioral data are standardized to ensure that the data of different dimensions have a consistent scale. Let the emotional data be E and the behavioral data be B. The standardization formula is: and Among them, μ E and σ E are the mean and standard deviation of sentiment data, μ B and σ B are the mean and standard deviation of the behavioral data, respectively; Feature extraction: Use principal component analysis to extract features from the standardized data to reduce the data dimension and retain the main information. The calculation formula is: Z = XW, where X is the standardized data matrix, W is the weight matrix containing the principal component direction, and Z is the feature matrix after dimensionality reduction; Emotional evaluation: Based on the extracted emotional features, a multidimensional sentiment analysis model is used to evaluate the students’ emotional state. The emotional evaluation formula is: Emotion score = α·AttentionScore+β·EngagementScore+γ·EmotionScore; where α, β and γ are weight coefficients, corresponding to the scores of concentration, engagement and emotional fluctuations respectively; Behavior assessment: Based on the extracted behavioral characteristics, a multidimensional behavior analysis model is used to evaluate students’ learning behavior. The behavior assessment formula is: Among them, w i is the weight of the i-th behavior feature, BehaviorFeature i is the value of the i-th behavior feature; Comprehensive evaluation: The emotional score and behavioral score are comprehensively analyzed to generate evaluation indicators and total scores for each dimension. The comprehensive evaluation formula is: comprehensive score = λ·emotional score + δ·behavioral score, where λ and δ are the weight coefficients of the emotional score and behavioral score.

10. The microteaching effectiveness evaluation system based on artificial intelligence according to claim 1 is characterized in that: The personalized teaching improvement suggestion module includes a data receiving unit, a feature extraction unit, a model training unit, a strategy generation unit and a resource recommendation unit; wherein: Data receiving unit: receiving the comprehensive evaluation report provided by the intelligent evaluation generation module, which includes evaluation indicators and scores in multiple dimensions of teaching quality, student participation and knowledge mastery; Feature extraction unit: extracts features from the evaluation report, including the teacher's teaching behavior, students' emotional feedback and learning behavior patterns, to generate personalized teaching improvement suggestions; Model training unit: Using historical teaching data and corresponding teaching effect evaluation, the machine learning model is trained through supervised learning methods. The model training unit includes three steps: data preprocessing, feature selection, and model training, and finally generates a model that can predict the best teaching strategies and resources; Strategy generation unit: Based on the trained machine learning model, it analyzes the characteristics of the current teaching evaluation report and generates specific teaching strategy suggestions, including teaching method adjustment, classroom interaction optimization and time management strategy; Resource recommendation unit: Use collaborative filtering algorithms to recommend appropriate teaching resources based on the needs of teachers and students. Teaching resources include teaching videos, courseware, exercise banks, and reference materials.

Images