Classroom effect evaluation and analysis method based on AI multi-mode six dimensions

Abstract

The invention discloses a multi-modal six-dimensional classroom effect evaluation and analysis method using AI, and the method comprises the steps: obtaining six-dimensional historical classroom data based on multi-modal historical classroom data; constructing an influence factor-weight mapping table; training a multi-modal fusion neural network model, performing feature fusion on the six-bit data, and constructing a classroom effect evaluation model; obtaining an influence factor value based on the real-time classroom time, and obtaining a weight distribution scheme with the highest matching degree with the current influence factor value; acquiring a classroom effect score, and performing targeted improvement based on the difference between the score and the optimal classroom effect score; and based on correction feedback data of students and teachers on the evaluation result, calculating the difference degree between model output and manual annotation through a contrast learning algorithm, and triggering model increment training. The method has the advantages that the weight is automatically adjusted by fusing multi-modal data and influence factors, the classroom effect is accurately evaluated, and personalized optimization suggestions are provided.

Description

Technical Field

[0001] The present invention relates to multi-modal data processing, and particularly to a method for evaluating and analyzing classroom effects using AI multi-modal six dimensions. Background Art

[0002] With the continuous development of information technology, especially the application of big data and artificial intelligence, the evaluation of classroom effects is no longer limited to traditional manual records and feedback analysis, but gradually develops towards intelligence and precision. By scientifically evaluating teaching activities and analyzing teaching effects, it provides a basis for teachers and education administrators to improve teaching quality. These methods usually include quantitative and qualitative analysis, and common tools are questionnaire surveys, classroom observations, student feedback, learning achievement analysis, etc. The background art of classroom effect evaluation involves the application of modern technologies such as data analysis, artificial intelligence, and learning analysis.

[0003] Current classroom effect evaluation methods on the market are often relatively single, usually relying only on data from a certain dimension, such as students' exam scores, classroom interaction data, or teachers' teaching feedback. These methods cannot comprehensively consider various factors inside and outside the classroom, such as environmental parameters, students' physiological signals, and real-time dynamic changes in the classroom, resulting in the evaluation results lacking depth and precision. In addition, many traditional evaluation methods rely on fixed evaluation criteria and manually set weights, making it difficult to flexibly adapt to different classroom situations or seasonal changes, and unable to provide personalized and real-time optimization suggestions. Moreover, existing evaluation systems usually lack the ability of self-correction and incremental learning, and it is difficult to automatically adjust the model according to real-time data and user feedback, increasing the error and limitations of the evaluation. Summary of the Invention

[0004] In order to improve the existing classroom effect evaluation method, a method for evaluating and analyzing classroom effects using AI multi-modal six dimensions is provided. This method automatically adjusts weights by integrating multi-modal data and influencing factors, accurately evaluates classroom effects, and provides personalized optimization suggestions; the real-time data input and feedback correction mechanism enables the model to continuously optimize itself, improving the accuracy and efficiency of the teaching process.

[0005] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for evaluating and analyzing classroom effects using AI multi-modal six dimensions, comprising:

[0007] Based on multi-modal historical classroom data, obtain historical classroom data in six dimensions, including visual data, voice data, physiological signal data, environmental parameter data, text interaction data, and student behavior trajectory data;

[0008] Construct an influence factor - weight mapping table based on time factors, exam activity factors, and seasonal weather factors;

[0009] Based on the acquired multi - modal historical classroom data, train a multi - modal fusion neural network model, perform feature fusion on six - dimensional data, and construct a classroom effect evaluation model. The classroom effect evaluation model includes a classroom effect evaluation model corresponding to the weight distribution scheme for achieving the best classroom effect under each influence factor;

[0010] Obtain the influence factor values based on the real - time classroom time, and obtain the weight distribution scheme with the highest matching degree with the current influence factor values;

[0011] Based on the classroom effect evaluation model corresponding to the obtained weight distribution scheme, input the six - dimensional classroom data collected in real - time into it, obtain the classroom effect score, and based on the gap between the score and the best classroom effect score, make targeted improvements to the insufficient dimensions of the six - dimensional data;

[0012] Based on the correction feedback data of students and teachers on the evaluation results, calculate the difference between the model output and the manual annotation through a contrastive learning algorithm and trigger incremental training of the model.

[0013] Preferably, the real - time acquisition of six - dimensional classroom data through a multi - modal sensor array includes: visual data, voice data, physiological signal data, environmental parameter data, text interaction data, and student behavior trajectory data, specifically including:

[0014] Obtain classroom teaching video data and voice data through high - resolution cameras and depth sensors, and convert classroom discussions, questions, and answers in real - time;

[0015] Obtain the physiological reactions of students and teachers through physiological signal collection devices worn by students and teachers to reflect emotion and attention information;

[0016] Real - time monitor the classroom physical environment through sensor devices, including temperature, humidity, air quality, and light intensity data;

[0017] Record students' writing, question - asking, and answering interaction data through the classroom interaction system;

[0018] Real - time track the head postures of students through the YOLOv7 + DeepSORT algorithm, capture the behavior actions of students and teachers, and construct a three - dimensional limb movement model.

[0019] Preferably, the construction of the influence factor - weight mapping table based on time factors, exam activity factors, and seasonal weather factors specifically includes:

[0020] For time factors, according to the classroom data in each time period, obtain the attention and participation data of students in different time periods;

[0021] For exam activity factors, based on classroom data related to exams, obtain data on students' exam stress, learning attitudes, and participation levels;

[0022] For seasonal weather factors, based on classroom data under different seasons and weather conditions, obtain data on students' emotions, behaviors, and learning effects;

[0023] Based on the impact of each influencing factor on students, construct an influencing factor - weight mapping table to store the weight allocation rules for each influencing factor under different circumstances.

[0024] Preferably, based on the obtained multi - modal historical classroom data, train a multi - modal fusion neural network model to perform feature fusion on six - dimensional data and construct a classroom effect evaluation model. The classroom effect evaluation model specifically includes the weight allocation schemes corresponding to achieving the best classroom effect under each influencing factor, and specifically includes:

[0025] Perform feature extraction based on the obtained multi - modal historical classroom data;

[0026] Fuse multi - modal features by training a multi - modal fusion neural network model;

[0027] Early fusion combines features of different modalities and inputs them into the neural network at the input stage;

[0028] Late fusion processes the data of each modality separately through independent neural networks and finally fuses them at the output layer, combining the outputs of each modality to obtain the final classroom effect evaluation result;

[0029] Perform deep fusion by designing multiple levels of fusion structures inside the neural network;

[0030] Based on the training results of the multi - modal fusion neural network, construct a classroom effect evaluation model, and based on the combined training of various influencing factors on the model, generate a corresponding weight allocation scheme.

[0031] Preferably, the method for obtaining the influencing factor values based on the real - time classroom time and obtaining the weight allocation scheme with the highest matching degree to the current influencing factor values specifically includes:

[0032] Obtain the classroom time and convert it into time factor data, including the current time period of the class, day of the week, and semester cycle;

[0033] Obtain the time from the current period to the exam or activity and the classroom participation and emotion data of students before and after the exam, and convert them into exam activity factor data;

[0034] Obtain the seasonal weather change data for the current period and convert it into seasonal weather factor data;

[0035] Based on the constructed influence factor - weight mapping table, the time factor data, exam activity factor data, and seasonal weather factor data are respectively automatically compared with the preset mapping table to obtain a set of allocation schemes with the highest data matching degree;

[0036] Based on the allocation scheme with the highest matching degree, obtain the weight values of each influence factor and the corresponding classroom effect evaluation model.

[0037] Preferably, for the classroom effect evaluation model corresponding to the obtained weight allocation scheme, input the real - time collected six - dimensional classroom data into it to obtain a classroom effect score, and based on the gap between the score and the best classroom effect score, specifically improve the insufficient dimensions in the six - dimensional data, including:

[0038] Based on the obtained classroom effect evaluation model, input the real - time collected six - dimensional classroom data into it to obtain a classroom effect score result;

[0039] Based on the comparison between the obtained classroom effect score result and the best classroom effect score of the classroom effect evaluation model, obtain the total score difference and the score differences of each dimension data;

[0040] Based on the total score difference and the score differences of each dimension data, improve and supplement the dimensions with low scores.

[0041] Preferably, for the correction feedback data of students and teachers on the evaluation results, calculate the difference between the model output and the manual annotation through a contrast learning algorithm and trigger model incremental training, specifically including:

[0042] Based on the correction feedback of students and teachers on the classroom effect evaluation obtained after the class, associate each correction feedback data with the actual classroom effect score and annotate it to obtain the difference between the annotated original score and the corrected score;

[0043] Based on the samples with large differences in the correction feedback data as the samples for incremental training, continuously adjust and optimize the model parameters through model incremental training.

[0044] Compared with the prior art, the advantages of the present invention are:

[0045] Comprehensively evaluate the classroom effect through multi-modal fusion, using data from six dimensions including vision, speech, physiological signals, environmental parameters, text interaction, and student behavior trajectories, providing a rich and multi-level perspective for classroom evaluation. By constructing an influencing factor-weight mapping table and combining variables such as time, exam activities, and seasonal weather, the model can automatically adjust the evaluation strategy according to different situations, thereby optimizing the classroom effect. In addition, by adopting the method of real-time data input and dynamic adjustment, the classroom effect evaluation can quickly respond according to specific classroom conditions and changes, ensuring the timeliness and accuracy of the evaluation results. Through a corrective learning mechanism based on the feedback of students and teachers, the model can continuously self-improve and incrementally train, further improving the prediction accuracy. This method not only realizes the refined and personalized evaluation of the classroom effect, but also provides targeted feedback and improvement plans for teachers and students, helping to achieve a more efficient teaching process and student learning experience. Brief Description of the Drawings

[0046] Figure 1 Schematic diagram of the method proposed by the present invention;

[0047] Figure 2 Schematic diagram of data collection proposed by the present invention;

[0048] Figure 3 Schematic diagram of the construction of the influencing factor-weight mapping table proposed by the present invention;

[0049] Figure 4 Schematic diagram of the construction of the classroom effect evaluation model proposed by the present invention;

[0050] Figure 5 Schematic diagram of model matching proposed by the present invention;

[0051] Figure 6 Schematic diagram of the classroom effect evaluation scoring proposed by the present invention;

[0052] Figure 7 Schematic diagram of corrective feedback proposed by the present invention;

[0053] Figure 8 Architectural diagram of the electronic device in this solution;

[0054] Figure 9 Schematic diagram of the structure of the computer-readable storage medium in this solution. Detailed Embodiment

[0055] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations.

[0056] Refer to Figure 1As shown in the figure, a method for evaluating and analyzing classroom effects using AI multi-modal six dimensions includes:

[0057] Step 1: Based on multi-modal historical classroom data, obtain historical classroom data in six dimensions, including visual data, speech data, physiological signal data, environmental parameter data, text interaction data, and student behavior trajectory data;

[0058] Step 2: Construct an influence factor-weight mapping table based on time factors, exam activity factors, and seasonal weather factors;

[0059] Step 3: Based on the obtained multi-modal historical classroom data, train a multi-modal fusion neural network model to perform feature fusion on the six-dimensional data and construct a classroom effect evaluation model. The classroom effect evaluation model includes a classroom effect evaluation model corresponding to the weight allocation scheme for achieving the best classroom effect under each influence factor;

[0060] Step 4: Obtain the influence factor value based on the real-time classroom time and obtain the weight allocation scheme with the highest matching degree with the current influence factor value;

[0061] Step 5: Based on the classroom effect evaluation model corresponding to the obtained weight allocation scheme, input the real-time collected six-dimensional classroom data into it to obtain the classroom effect score, and based on the gap between the score and the best classroom effect score, make targeted improvements to the insufficient dimensions in the six-dimensional data;

[0062] Step 6: Based on the correction feedback data of students and teachers on the evaluation results, calculate the difference between the model output and the manual annotation through a contrastive learning algorithm and trigger incremental training of the model.

[0063] Refer to Figure 2 As shown in the figure, six-dimensional classroom data is collected in real time through a multi-modal sensor array, including: visual data, speech data, physiological signal data, environmental parameter data, text interaction data, and student behavior trajectory data specifically includes:

[0064] Obtain classroom teaching video data and speech data through high-resolution cameras and depth sensors, and convert classroom discussions, questions, and answers in real time;

[0065] Obtain the physiological reactions of students and teachers through physiological signal collection devices worn by students and teachers to reflect emotions and attention information;

[0066] Real-time monitor the classroom physical environment through sensor devices, including temperature, humidity, air quality, and light intensity data;

[0067] Record students' writing, question, and answer interaction data through the classroom interaction system;

[0068] Track the head postures of students in real time through the YOLOv7+DeepSORT algorithm, capture the behavior actions of students and teachers, and build a three-dimensional limb action model.

[0069] Refer to Figure 3 As shown, the construction of the influence factor-weight mapping table based on time factors, exam activity factors, and seasonal weather factors specifically includes:

[0070] For time factors, according to the classroom data of each time period, obtain the attention and participation data of students in different time periods;

[0071] For exam activity factors, according to the classroom data related to exams, obtain the exam stress, learning attitude, and participation data of students;

[0072] For seasonal weather factors, according to the classroom data under different seasons and weather conditions, obtain the student mood, behavior, and learning effect data;

[0073] Based on the influence of each influencing factor on students, build an influence factor-weight mapping table to store the weight distribution rules of each influencing factor under different circumstances.

[0074] Specifically, when allocating weights for time factors, different weights are allocated according to the performance of students in different time periods. For example, a higher weight may be allocated in the morning time period because students' attention and learning status are usually better. The participation and attention of students change over time, and the following model can be used for quantification. The formula is:

[0075] A t =α1T t +α2S t +α3E t

[0076] Among them, A t is the attention and participation of students in time period t, T t is the time factor, such as morning, afternoon, before and after class, S t is the physiological signal of students, such as heart rate, brain waves, etc., E t is the environmental factor, and α1, α2, α3 are the weight coefficients of each factor in the time period;

[0077] When allocating weights for exam activities, weights are allocated according to the classroom performance before and after the exam. When approaching the exam, students may be more concerned about the exam and the classroom participation may be lower. Therefore, the weight of classroom effect can be reduced; the seasonal weather weight is allocated based on the changes in weather and seasons. For example, summer or cold winter may affect the learning mood of students, so the weight of classroom effect needs to be adjusted;

[0078] Based on the above content, a weight mapping table for each factor is constructed through different data analysis methods (such as regression analysis and principal component analysis). Among the time factors, students' attention is lower in the morning and higher in the afternoon. The weight settings are shown in the following table:

[0079] Influencing factor Time period Weight coefficient Time factor Morning <![CDATA[α1 = 0.2, α2 = 0.3, α3 = 0.5]]> Time factor Afternoon <![CDATA[α1 = 0.4, α2 = 0.3, α3 = 0.3]]>

[0080] The weight allocation of data in each dimension can be dynamically adjusted through a reinforcement learning algorithm. The reinforcement learning algorithm interacts with classroom data and continuously adjusts decisions (i.e., weight allocation) according to environmental feedback, and finally learns the optimal weight allocation scheme to obtain the best classroom effect.

[0081] Refer to Figure 4 As shown, based on the obtained multi-modal historical classroom data, a multi-modal fusion neural network model is trained to perform feature fusion on six-dimensional data and construct a classroom effect evaluation model. The classroom effect evaluation model specifically includes the weight allocation scheme corresponding to the best classroom effect under each influencing factor:

[0082] Feature extraction is performed based on the obtained multi-modal historical classroom data;

[0083] The multi-modal features are fused through training a multi-modal fusion neural network model;

[0084] Early fusion combines features of different modalities at the input stage and inputs them into the neural network;

[0085] Late fusion processes the data of each modality through independent neural networks respectively, and finally performs fusion at the output layer. The outputs of each modality are combined to obtain the final classroom effect evaluation result;

[0086] Deep fusion is performed by designing a fusion structure with multiple levels inside the neural network;

[0087] Based on the training results of the multi-modal fusion neural network, a classroom effect evaluation model is constructed, and a corresponding weight allocation scheme is generated based on the trained model under various combinations of influencing factors.

[0088] Specifically, in early fusion, features of different modalities are combined at the input stage to form a unified feature vector, and the combined feature vector is input into the first layer of the neural network for training. In late fusion, the features of each modality are processed through independent neural networks respectively, and then fused at the output layer to obtain the final classroom effect evaluation result. The formula is:

[0089] Y final =σ(T t ·W video +T t ·Wvoice +T t ·E t ·S t ·W feel +T t ·E t ·S t ·W env +T t ·W text

[0090] +T t ·E t ·S t ·W move )

[0091] Among them, T t , E t [[ID=]38], S t are the weight of time factor, the weight of examination activity factor, and the weight of season and weather factor respectively. W video is visual data, W voice is voice data, W feel is physiological signal data, W env is environmental parameter data, W text is text interaction data, W move is student behavior trajectory data;

[0092] By training a deep learning model, the weights of each modality and the fusion layer are obtained. The loss function used in the training process is usually a regression loss function. Based on the training results of the multi-modal fusion neural network, a weight allocation scheme is constructed. The influence degree of each modality in the final classroom effect evaluation is determined by the weight of the modality, and the weights are dynamically adjusted according to different situations (such as time period, examination, season, etc.).

[0093] Refer to Figure 5 As shown, obtaining the influence factor values based on the real-time classroom time and obtaining the weight allocation scheme with the highest matching degree with the current influence factor values specifically include:

[0094] Obtain the classroom time and convert it into time factor data, including the current time period of the classroom, the week date, and the semester cycle;

[0095] Obtain the time from the current period to the examination or activity and the classroom participation and emotion data of the students before and after the examination, and convert them into examination activity factor data;

[0096] Obtain the seasonal weather change data of the current period and convert it into seasonal weather factor data;

[0097] Based on the constructed influence factor - weight mapping table, compare the time factor data, exam activity factor data, and seasonal weather factor data with the preset mapping table respectively to obtain a set of allocation schemes with the highest data matching degree;

[0098] Based on the allocation scheme with the highest matching degree, obtain the weight values of each influence factor and the corresponding classroom effect evaluation model among them.

[0099] Specifically, compare the time factor data, exam activity factor data, and seasonal weather factor data with the preset influence factor - weight mapping table, and select the allocation scheme that best matches the current classroom state. The formula is:

[0100]

[0101] Among them, X i is the actual data of the current time period, and w i is the weight value in the preset mapping table.

[0102] Refer to Figure 6 As shown, based on the classroom effect evaluation model corresponding to the obtained weight allocation scheme, input the real - time collected six - dimensional classroom data into it to obtain the classroom effect score, and based on the gap between the score and the best classroom effect score, make targeted improvements to the insufficient dimensions in the six - dimensional data. Specifically include:

[0103] Based on the obtained classroom effect evaluation model, input the real - time collected six - dimensional classroom data into it to obtain the classroom effect score result;

[0104] Based on the comparison between the obtained classroom effect score result and the best classroom effect score of the classroom effect evaluation model, obtain the total score difference and the score differences of each dimension data;

[0105] Based on the total score difference and the score differences of each dimension data, improve and supplement the dimensions with low scores.

[0106] Specifically, the total score difference is the difference between the classroom effect score and the best classroom effect score. The formula is:

[0107] ΔY total =|Y class -Y best |

[0108] Among them, ΔY total is the total score difference, Y class is the classroom effect score, and Y best is the best classroom effect score;

[0109] Calculate the score difference of each dimension based on the data of each dimension. The formula is:

[0110]

[0111] Among them, Y i is the score of the current dimension, and is the score of this dimension under the best classroom effect;

[0112] When we identify which dimensions have low scores, we can take different improvement measures according to the specific situation of each dimension, including:

[0113] Emotion dimension: Reduce anxiety and improve the emotional state by adjusting the classroom atmosphere and giving students more emotional support;

[0114] Engagement dimension: Increase classroom interaction, such as group discussions, question-and-answer sessions, etc., to promote students' more active participation;

[0115] Attention dimension: Improve students' concentration by improving the classroom environment (such as lighting, temperature, etc.) or designing more attractive teaching activities;

[0116] Physiological data: Improve students' physical state by reducing long periods of sitting and arranging more rest time.

[0117] Refer to Figure 7 As shown, based on the corrected feedback data of students and teachers on the evaluation results, calculating the difference between the model output and the manual annotation through a contrast learning algorithm and triggering model incremental training specifically includes:

[0118] Based on the corrected feedback obtained from students and teachers on the classroom effect evaluation after the class, associate each corrected feedback data with the actual classroom effect score and annotate it to obtain the difference between the annotated original score and the corrected score;

[0119] Use the samples with large differences in the corrected feedback data as samples for incremental training, and continuously adjust and optimize the model parameters through model incremental training.

[0120] Specifically, after the class, students and teachers provide corrected opinions on the classroom effect through questionnaires, classroom evaluation systems or other feedback mechanisms. Student feedback includes students' evaluations of classroom content, teaching methods, interaction forms, etc., as well as their subjective scores on the classroom effect. Teacher feedback includes teachers' self-evaluations of the classroom effect, including aspects such as teaching methods, student participation, and classroom management;

[0121] Associate the corrected feedback data of students and teachers with the actual classroom effect score (the score generated based on real-time data) to create a dataset, where each corrected feedback is associated with the corresponding actual score, and annotate the difference between the original score and the corrected score. Each record includes the original score, the corrected score, the score difference, and the corrected feedback data;

[0122] Based on the samples with large score differences in the dataset as incremental training samples, on the basis of the existing model, through the newly obtained data for gradual learning and updating, continuously adjust and optimize the model parameters, adjust the model parameters through incremental data, so that the classroom effect score predicted by the model is as close as possible to the corrected score. The objective function formula is:

[0123]

[0124] Among them, f(X i ,θ) is the classroom effect score, X i is the feature data, θ is the parameter, is the corrected score;

[0125] As more corrected feedback data is continuously input, the model can be continuously adjusted through incremental training to improve the accuracy and generalization ability of the model. After each incremental training, the performance of the model will be evaluated through a new test set, and the model will be further optimized as needed.

[0126] Furthermore, the method according to the embodiment of the present application can also be implemented by means of Figure 8 the architecture of the electronic device shown. As Figure 8 shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to the network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, can store a method for evaluating and analyzing the classroom effect using AI multi-modal six dimensions provided by the present application. The electronic device 500 may further include a terminal interface 508. Of course, Figure 8 the architecture shown is only exemplary. When implementing different devices, one or more components in the electronic device shown may be omitted according to actual needs. Figure 8

[0127] Figure 9 Figure 9 is a schematic diagram of the structure of a computer-readable storage medium provided by an embodiment of the present application. As Figure 9As shown, it is a computer-readable storage medium 600 according to an embodiment of the present application. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are run by a processor, a method for evaluating and analyzing classroom effects using AI multi-modal six dimensions according to an embodiment of the present application described with reference to the above drawings can be executed. The storage medium 600 includes but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory, etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0128] It should be noted that: the above sequence of the embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. And the above specific embodiments of this specification have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0129] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the differences between each embodiment and other embodiments are emphasized.

[0130] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating and analyzing classroom effects using AI multi-modal six dimensions, characterized in that, Including: Based on multi-modal historical classroom data, obtain historical classroom data in six dimensions, including visual data, voice data, physiological signal data, environmental parameter data, text interaction data, and student behavior trajectory data; Construct an influence factor-weight mapping table based on time factors, exam activity factors, and seasonal weather factors; Based on the obtained multi-modal historical classroom data, train a multi-modal fusion neural network model, perform feature fusion on the six-dimensional data, and construct a classroom effect evaluation model. The classroom effect evaluation model includes a classroom effect evaluation model corresponding to the weight allocation scheme for achieving the best classroom effect under each influence factor; Obtain the influence factor values based on the real-time classroom time, and obtain the weight allocation scheme with the highest matching degree with the current influence factor values; Based on the classroom effect evaluation model corresponding to the obtained weight allocation scheme, input the six-dimensional classroom data collected in real time into it, obtain the classroom effect score, and based on the gap between the score and the best classroom effect score, make targeted improvements to the insufficient dimensions in the six-dimensional data; Based on the correction feedback data of students and teachers on the evaluation results, calculate the difference between the model output and the manual annotation through the contrast learning algorithm and trigger incremental training of the model.

2. The method for evaluating and analyzing classroom effects using AI multi-modal six dimensions according to claim 1, wherein The real-time collection of six-dimensional classroom data through a multi-modal sensor array includes: Visual data, voice data, physiological signal data, environmental parameter data, text interaction data, and student behavior trajectory data specifically include: Obtain classroom teaching video data and voice data through a high-resolution camera and a depth sensor, and convert classroom discussions, questions, and answers in real time; Obtain the physiological reactions of students and teachers through physiological signal collection devices worn by students and teachers, reflecting emotions and attention information; Real-time monitor the classroom physical environment through sensor devices, including temperature, humidity, air quality, and light intensity data; Record students' writing, question, and answer interaction data through the classroom interaction system; Real-time track the head postures of students through the YOLOv7+DeepSORT algorithm, capture the behavior actions of students and teachers, and construct a three-dimensional limb movement model.

3. A method for evaluating and analyzing classroom effects using AI multi-modal six dimensions according to claim 1, characterized in that, The construction of an influence factor-weight mapping table based on time factors, exam activity factors, and seasonal weather factors specifically includes: For time factors, according to the classroom data in each time period, obtain the attention and participation data of students in different time periods; For exam activity factors, according to the classroom data related to exams, obtain the exam pressure, learning attitude, and participation data of students; For seasonal weather factors, according to the classroom data in different seasons and weather conditions, obtain the emotions, behaviors, and learning effect data of students; Based on the influence of each influence factor on students, construct an influence factor-weight mapping table, and store the weight allocation rules of each influence factor in different situations.

4. A method for evaluating and analyzing classroom effects using AI multi-modal six dimensions according to claim 1, characterized in that, The training of a multi-modal fusion neural network model based on the obtained multi-modal historical classroom data, performing feature fusion on the six-dimensional data, and constructing a classroom effect evaluation model. The classroom effect evaluation model includes a classroom effect evaluation model corresponding to the weight allocation scheme for achieving the best classroom effect under each influence factor specifically includes: Feature extraction is performed based on the acquired multi-modal historical classroom data; The multi-modal features are fused by training a multi-modal fusion neural network model; Early fusion combines features of different modalities and inputs them into the neural network at the input stage; Late fusion processes the data of each modality through independent neural networks respectively, and finally fuses them at the output layer, combining the outputs of each modality to obtain the final classroom effect evaluation result; Deep fusion is performed by designing a fusion structure with multiple levels inside the neural network; Based on the training results of the multi-modal fusion neural network, a classroom effect evaluation model is constructed, and the trained model is combined with various influencing factors to generate a corresponding weight allocation scheme.

5. A method for evaluating and analyzing classroom effects using AI multi-modal six dimensions according to claim 1, characterized in that The method for obtaining the influencing factor values based on the real-time classroom time and obtaining the weight allocation scheme with the highest matching degree with the current influencing factor values specifically includes: Obtain the classroom time and convert it into time factor data, including the current time period of the class, week date, and semester cycle; Obtain the time from the current period to the exam or activity and the classroom participation and emotion data of students before and after the exam, and convert them into exam activity factor data; Obtain the seasonal weather change data of the current period and convert it into seasonal weather factor data; Based on the constructed influencing factor-weight mapping table, compare the time factor data, exam activity factor data, and seasonal weather factor data with the preset mapping table respectively to obtain a set of allocation schemes with the highest data matching degree; Based on the allocation scheme with the highest matching degree, obtain the weight values of each influencing factor and the corresponding classroom effect evaluation model.

6. A method for evaluating and analyzing classroom effects using AI multi-modal six dimensions according to claim 1, characterized in that, Based on the classroom effect evaluation model corresponding to the obtained weight allocation scheme, input the real-time collected six-dimensional classroom data into it to obtain the classroom effect score, and based on the gap between the score and the best classroom effect score, make targeted improvements to the insufficient dimensions in the six-dimensional data. Specifically include: Based on the obtained classroom effect evaluation model, input the real-time collected six-dimensional classroom data into it to obtain the classroom effect score result; Compare the obtained classroom effect score result with the best classroom effect score of the classroom effect evaluation model to obtain the total score difference and the score differences of each dimension data; Based on the total score difference and the score differences of each dimension data, improve and supplement the dimensions with low scores.

7. A method for evaluating and analyzing classroom effects using AI multi-modal six dimensions according to claim 1, characterized in that, Based on the correction feedback data of students and teachers on the evaluation results, calculate the difference between the model output and the manual annotation through the contrast learning algorithm and trigger the incremental training of the model. Specifically include: Based on the correction feedback of students and teachers on the classroom effect evaluation obtained after the class, associate each correction feedback data with the actual classroom effect score and annotate it to obtain the difference between the annotated original score and the corrected score; Use the samples with large differences in the correction feedback data as the samples for incremental training, and continuously adjust and optimize the model parameters through model incremental training.

8. An electronic device, characterized in that, Including: At least one processor; And a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a method for classroom effect evaluation and analysis using AI multi-modal six dimensions as described in any one of claims 1-7.

9. A computer-readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by a processor, a method for classroom effect evaluation and analysis using AI multi-modal six dimensions as described in any one of claims 1-7 is implemented.

Images