AI-based art teaching management system and method thereof

Abstract

The invention discloses an art teaching management system and method based on AI. The system comprises an intelligent user management module, a course generation module, a real-time interaction module, an evaluation module, an interaction control module and a data security module. Through a multi-modal deep integration learning algorithm MMDILA and a reinforcement learning algorithm, multi-modal data of a user is analyzed, dynamic generation and adjustment of personalized course content are realized, and real-time learning feedback and personalized suggestions are provided. And a block chain technology and a differential privacy technology are adopted to guarantee the security and privacy of user data. According to the method, the personalized teaching effect and learning efficiency are remarkably improved, the learning time is shortened, the user satisfaction is enhanced, the data security and the privacy protection capability are greatly improved, and the defects in the aspects of personalization, real-time feedback and data security in the prior art are overcome. The method is obviously superior to the prior art in the aspects of personalized teaching and data security, and can meet the requirements of different users.

Description

Technical Field

[0001] The present invention relates to the field of chip testing technology, and in particular to an AI-based art teaching management system and method thereof. Background Art

[0002] In traditional art teaching management systems, the commonly used methods are mainly based on fixed learning content and linear teaching path course design. These systems usually lack the ability to analyze and adjust individual user learning behaviors, progress and performance in real time, making it difficult for teaching content to dynamically adapt to user learning needs and interest changes. In addition, most systems lack effective user data security protection mechanisms, which are prone to privacy leakage and data security risks when processing and storing user data.

[0003] Some art teaching management systems have the following major problems: First, these systems usually use static and predefined learning paths, lack the ability to respond to users' personalized needs, and cannot adjust course content and teaching strategies based on users' specific performance and feedback. Secondly, the interaction and feedback mechanisms in these systems are limited, and they cannot capture and process users' multimodal learning data (such as text, images, voice, etc.) in real time, thus affecting the effectiveness of personalized teaching. In addition, existing systems have insufficient measures for data security, especially when it comes to user sensitive data, lack effective encryption, privacy protection and anomaly detection mechanisms, which may lead to data leakage and abuse. Summary of the invention

[0004] In view of the problems of limited personalized learning ability, lack of real-time interaction and feedback mechanism, and insufficient data security and privacy protection measures, the present invention proposes an AI-based art teaching management system and method thereof, which optimizes the functions of course generation, interactive feedback, and data security by introducing a variety of advanced AI technologies, such as deep learning, natural language processing, blockchain, and differential privacy. The present invention can dynamically analyze the user's multimodal learning data, automatically generate and adjust personalized course content, provide learning feedback in real time, and ensure the security and privacy of user data through a variety of data protection technologies (such as blockchain and differential privacy technology).

[0005] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0006] An AI-based art teaching management system, the system comprising:

[0007] An intelligent user management module is used to identify the user identity and dynamically adjust the allocation of learning resources and permission configuration using a support vector machine (SVM) algorithm based on the user's learning behavior data and historical records. The intelligent user management module is based on an improved multimodal deep integrated learning algorithm (MMDILA) and optimizes the data preprocessing and analysis process by performing unified modeling, feature selection and data fusion on multimodal data;

[0008] The course generation module generates personalized course content based on the user's learning progress and performance through the convolutional neural network (CNN) model and random forest algorithm;

[0009] The real-time interaction module analyzes the user's interactive content through convolutional neural network (CNN) and natural language processing (NLP) technology, and generates real-time learning feedback and personalized suggestions;

[0010] The evaluation module uses reinforcement learning algorithms to automatically evaluate user-submitted works and generate feedback;

[0011] Interactive control module, which uses speech recognition, gesture recognition and image recognition technologies to optimize the learning interface;

[0012] The data security module combines blockchain and differential privacy technology to protect user data and monitor data access behavior in real time.

[0013] Preferably, the intelligent user management module further includes:

[0014] The user identification unit uses a deep neural network (DNN) AI algorithm to efficiently and accurately identify users and provide personalized access rights control;

[0015] The data analysis unit uses multimodal learning data (including text, images, and voice) for comprehensive analysis to generate personalized learning paths and progress plans that meet user characteristics;

[0016] The resource allocation unit uses graph neural networks (GNNs) and Transformer models to perform deep modeling and sequence analysis on user data and dynamically adjust the allocation of learning resources;

[0017] The resource allocation unit further models and performs sequence analysis on user data through graph neural network (GNN) and Transformer model, and realizes resource optimization configuration by combining meta-learning and adaptive learning rate algorithm.

[0018] The specific implementation steps are as follows:

[0019] Step 1: Use multimodal learning combined with graph neural network (GNN) to model the relationship network between users, perform fusion analysis on the multimodal data of users, and generate user feature representation. The calculation formula is as follows:

[0020]

[0021] in, represents the feature representation of node v at the k+1th layer, σ is a nonlinear activation function, N(v) is the set of neighbor nodes of node v,

[0022] c vu is the normalization factor, W (k) is the learnable weight matrix of the kth layer;

[0023] Step 2: Use the Transformer model based on the self-attention mechanism to perform sequence modeling on the user's historical behavior data, generate a personalized learning path, and optimize the parameters through meta-learning and adaptive learning rate algorithm. The calculation formula is as follows:

[0024]

[0025] Among them, Q is the query matrix, K is the key matrix, and V is the value matrix. is the dimension of the key vector;

[0026] Through meta-learning and adaptive learning rate algorithms (such as Adagrad or RMSProp), the learning path is dynamically adjusted to optimize the parameter θ:

[0027]

[0028] Where Gt is the accumulated squared gradient, ∈ is the smoothing term;

[0029] Step 3: Use non-negative matrix factorization (NMF) to preliminarily model the user feature matrix and resource utility matrix, and further optimize the resource allocation strategy through diffusion models. The calculation formula is as follows:

[0030]

[0031] Among them, pθ(x t-1 ∣x t ) is the given current state x t Generate the previous state x t-1 The conditional probability, μθ(xt ,t) and is the mean and variance predicted by the neural network, and I is the identity matrix;

[0032] Step 4: Combine real-time sentiment analysis technology to dynamically adjust learning content and feedback expression. The calculation formula is as follows;

[0033]

[0034] Among them, Q(s,a) is the state-action value function, s is the current state, a is the current action, α is the learning rate, r is the immediate reward, γ is the discount factor, a′ is the possible future action, and s′ is the next state;

[0035] Step 5: Use Differential Privacy and Federated Learning methods to protect user data.

[0036] Preferably, the course generation module includes:

[0037] The content generation unit uses an improved convolutional neural network (CNN) and stochastic gradient descent (SGD) optimization algorithm to automatically generate diverse personalized course content based on user learning needs and interests;

[0038] Data fusion unit, which efficiently integrates multimodal learning data (text, images, videos), processes data and extracts features through deep learning technology, and generates course content that suits the user's learning style;

[0039] The dynamic adjustment unit uses deep reinforcement learning algorithms to intelligently adjust the course difficulty, teaching strategy and content presentation method based on the user's learning progress and real-time performance to adapt to changes in the user's learning status.

[0040] Preferably, the real-time interaction module includes:

[0041] The interactive analysis unit extracts interactive text features through word embedding method based on bidirectional long short-term memory network (Bi-LSTM) and convolutional neural network (CNN);

[0042] The emotion recognition unit fuses the extracted text features with the user’s voice and visual data through a multimodal emotion fusion network;

[0043] The feedback generation unit generates personalized learning feedback based on the Transformer model and dynamically adjusts the feedback strategy through the reinforcement learning algorithm.

[0044] Preferably, the evaluation module comprises:

[0045] A multi-dimensional evaluation unit, which is used to automatically evaluate and analyze user-submitted artworks in terms of composition, color, technique, and creativity based on computer vision technology and machine learning algorithms;

[0046] The adaptive evaluation engine unit uses the user's historical performance data and real-time feedback to dynamically adjust the weight of each evaluation dimension to achieve personalized evaluation standards;

[0047] The safety optimization unit is used to optimize the evaluation criteria and strategies using the deep reinforcement learning algorithm DQN.

[0048] Preferably, the interactive control module includes:

[0049] The recognition unit uses a deep neural network (DNN) model to recognize the user's voice, gesture, and image input and convert the instructions into operation commands;

[0050] The interface optimization unit adjusts the layout, content presentation order and visual effects of the learning interface in real time according to the recognition results.

[0051] Preferably, the data security module includes:

[0052] The encryption unit uses the Advanced Encryption Standard (AES-256) algorithm combined with distributed blockchain technology to encrypt and distribute user data to ensure data integrity and security;

[0053] Privacy protection unit, which protects user data based on differential privacy technology to prevent the leakage of sensitive information during data analysis;

[0054] The monitoring unit uses behavioral analysis algorithms and zero-knowledge proof methods to monitor and detect data access behavior in real time, identify abnormal access, and support multi-party collaboration based on federated learning.

[0055] An AI-based art teaching management method is applied to a computing device, and the method comprises the following steps:

[0056] Step S1: Collecting multimodal learning data of the user, including but not limited to text, image, voice and video data;

[0057] Among them, text data is segmented and parsed using natural language processing technology; image data is collected by an image sensor with a resolution of 256x256 pixels and saved in JPEG format; the voice data sampling rate is 16kHz and stored in WAV format; the video data frame rate is 30fps and is encoded in MP4 format;

[0058] Step S2: extracting and fusing the collected multimodal data through the multimodal deep integrated learning algorithm MMDILA, wherein a convolutional neural network is used to extract features from image data, a bidirectional long short-term memory network is used to perform semantic analysis on text data, and a sentiment recognition model is used to process speech data. The Transformer model with a self-attention mechanism is used to fuse information from different data modalities to generate a comprehensive learning profile of the user;

[0059] Step S3: Based on the user learning portrait, dynamically adjust the personalized course content and teaching strategy through the reinforcement learning algorithm, wherein the reinforcement learning algorithm is based on the policy gradient method, the learning rate is 0.01, the discount factor is 0.9, and the optimization goal is to maximize the user learning effect and satisfaction.

[0060] Preferably, the integration process of the multi-level deep learning model further comprises the following steps:

[0061] The integration process of the multi-level deep learning model further includes the following steps:

[0062] Step 1: Through the multi-task learning framework, the outputs of the CNN model and the Bi-LSTM model are jointly trained in the shared layer;

[0063] Step 2: Apply self-supervised learning technology to pre-train users’ multimodal interaction data;

[0064] Step 3: Use the self-attention mechanism to dynamically adjust the importance of different data modalities.

[0065] Preferably, the step of dynamically adjusting personalized course content and teaching strategies includes:

[0066] Step 1: Use the improved multimodal deep ensemble learning algorithm (MMDILA) and adversarial generative network (GAN) model to predict user behavior;

[0067] Step 2: Train the generator to simulate the user's potential behavior and train the discriminator to distinguish the generated behavior from the actual behavior;

[0068] Step 3: Based on the results of the above adversarial training, adjust the user's learning path and feedback mechanism in real time.

[0069] Terminology explanation:

[0070] Multimodal Deep Integrated Learning Algorithm MMDILA: refers to a deep learning method that can process multiple types of data (such as text, images, voice and video) at the same time. By integrating convolutional neural networks (CNN) for image feature extraction, natural language processing (NLP) for text analysis, and sentiment recognition models for voice processing, the features of various types of data are finally integrated to generate personalized learning paths and recommendations for users. The innovation of MMDILA lies in its deep integration of features from multiple data sources, which improves the system's ability to understand and respond to user learning behaviors.

[0071] Reinforcement learning algorithm: It is a machine learning method used to dynamically adjust user learning paths and teaching strategies. Through continuous interaction with users, feedback data is obtained to evaluate the effectiveness of the current strategy. The system sets an objective function and reward mechanism to guide the learning process to optimize in the direction of the best strategy. The specific implementation of this algorithm includes models such as Q-learning or deep Q network (DQN), which maximizes the user's learning effect and satisfaction by continuously optimizing the strategy.

[0072] Blockchain technology: used to ensure the security and transparency of user data. It records all data access and operation behaviors through distributed ledgers to ensure the immutability and traceability of data. The present invention adopts a consortium chain structure, combined with a multi-party consensus mechanism, to record and monitor the data access process and ensure that all data operations in the system are secure and verifiable. This technology is further integrated into the data security module and works in conjunction with differential privacy technology.

[0073] Differential privacy technology: It is a statistical method to protect the privacy of user data. It ensures that individual information is unidentifiable in data analysis by adding noise to the data set. This technology prevents the leakage of sensitive information during data analysis and model training. The present invention uses differential privacy technology to ensure that the output results do not leak the specific information of any specific user when analyzing and processing user data. The system controls the amount of noise added by setting a privacy budget to strike a balance between data availability and privacy protection.

[0074] User portrait generation: refers to the system generating comprehensive learning characteristics and behavior portraits of users through multimodal data fusion and machine learning technology. Specifically, the present invention uses the Transformer model to fuse multimodal data (text, image, voice), extract the user's learning habits, preferences and performance characteristics, and form a personalized user portrait as the basis for course content generation and adjustment.

[0075] Real-time interaction module: includes interaction analysis unit, emotion recognition unit and feedback generation unit, which is used to interact and provide feedback to users in real time through multimodal data (voice, gesture, text, etc.). The interaction analysis unit is responsible for real-time analysis of user input data, the emotion recognition unit analyzes the user's emotional state, and the feedback generation unit provides instant personalized learning suggestions and course adjustments based on the analysis results to improve learning effects and user experience.

[0076] Multi-dimensional evaluation unit, adaptive evaluation engine: used to comprehensively evaluate the user's learning effect based on multiple indicators (such as learning progress, accuracy, participation, etc.). The "adaptive evaluation engine" dynamically adjusts the learning path and course content based on the results of the multi-dimensional evaluation unit to achieve the best learning effect. The engine combines reinforcement learning algorithms to continuously learn and optimize evaluation criteria and strategies to ensure the accuracy and reliability of evaluation results.

[0077] Compared with the prior art, the present invention has the following beneficial effects:

[0078] 1. Significantly improve personalized teaching effects: The present invention significantly enhances the system's ability to respond to users' personalized learning needs by introducing a multimodal deep integrated learning algorithm (MMDILA) and an improved support vector machine (SVM) algorithm. By analyzing users' multimodal data (such as text, voice, images, etc.), the present invention can dynamically generate and adjust personalized course content to adapt to users' learning progress and performance in real time. This innovative method overcomes the defects of fixed learning paths in the prior art, making the teaching process more flexible and effective, and helping to improve learning effects and user experience.

[0079] 2. Optimize real-time interaction and feedback mechanism: The present invention realizes real-time interaction and feedback with users by utilizing convolutional neural network (CNN) and natural language processing (NLP) technology. The system can generate instant feedback and personalized suggestions based on the user's real-time input (such as voice commands, gesture operations, etc.), and automatically evaluate the user's learning effect through reinforcement learning algorithms. This mechanism not only enhances the interactivity and fun of teaching, but also greatly improves the efficiency of teaching, makes up for the lack of real-time feedback in the existing technology, and significantly improves the user's sense of participation and satisfaction.

[0080] 3. Significantly improve data security and privacy protection: The present invention innovatively combines blockchain technology and differential privacy technology to effectively protect the security and privacy of user data. Blockchain technology is used to record and monitor data access behavior to ensure the transparency and immutability of data operations; differential privacy technology is used to protect user sensitive data and prevent it from being leaked during data analysis and use. Compared with the existing technology, the present invention significantly improves the security of data storage and transmission, and has made breakthrough progress in anomaly detection and data access monitoring, which comprehensively solves the deficiencies of the existing technology in data security protection.

[0081] 4. Innovative multimodal data fusion technology: This invention can efficiently fuse and process learning information from multiple data sources (such as user voice, image and text input) through an improved multimodal deep integrated learning algorithm (MMDILA) and convolutional neural network (CNN). This innovative data fusion method significantly improves the accuracy and speed of data analysis, enabling the system to understand users' learning habits and preferences more quickly and accurately, and generate more accurate personalized learning paths. This improvement on existing data processing methods significantly enhances the intelligence level and adaptive capabilities of the system.

[0082] 5. Improve the expansibility and compatibility of the system: The present invention adopts a modular system architecture design, and each functional module (such as user management, course generation, interactive feedback, data security, etc.) is independent and coordinated with each other, so that the system has high expansibility and compatibility in subsequent development and application. Compared with the prior art, this innovative architecture design not only facilitates the expansion and update of functions, but also reduces the maintenance cost of the system, which is conducive to wider application and promotion, and improves the market competitiveness and commercialization potential of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0084] in:

[0085] Figure 1 is a system architecture diagram of an embodiment of the present invention;

[0086] Figure 2 This is a schematic diagram of an intelligent user management module in an embodiment of the present invention;

[0087] Figure 3 This is a schematic diagram of a course generation module in an embodiment of the present invention;

[0088] Figure 4 A schematic diagram of a real-time interactive module in an embodiment of the present invention;

[0089] Figure 5 This is a schematic diagram of an evaluation module in an embodiment of the present invention;

[0090] Figure 6 The figure is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0091] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0092] Example 1

[0093] like Figure 1 As shown in FIG. 1 , an embodiment of the present invention is provided, which provides an AI-based art teaching management system, including:

[0094] An intelligent user management module is used to identify the user identity and dynamically adjust the allocation of learning resources and permission configuration using a support vector machine (SVM) algorithm based on the user's learning behavior data and historical records. The intelligent user management module is based on an improved multimodal deep integrated learning algorithm (MMDILA) and optimizes the data preprocessing and analysis process by performing unified modeling, feature selection and data fusion on multimodal data;

[0095] The core function of the intelligent user management module is to identify the user and dynamically adjust the allocation of learning resources and permission configuration based on the user's learning behavior data and history. This module uses the support vector machine (SVM) algorithm to analyze the user's behavior data to ensure that the system can flexibly respond to the different needs of users.

[0096] This module is based on the improved multimodal deep integrated learning algorithm (MMDILA), which optimizes the data preprocessing and analysis process by unified modeling, feature selection and data fusion of multimodal data (such as text, image, speech, etc.). The improvement of MMDILA is that it integrates the features of multiple data types, enabling the system to understand the user's learning behavior and preferences more accurately, thereby significantly optimizing the existing technology in terms of the diversity and complexity of data input.

[0097] Specifically, Figure 2 As shown, the intelligent user management module further includes:

[0098] The user identification unit uses a deep neural network (DNN) AI algorithm to efficiently and accurately identify users and provide personalized access rights control;

[0099] The data analysis unit uses multimodal learning data (including text, images, and voice) for comprehensive analysis to generate personalized learning paths and progress plans that meet user characteristics;

[0100] The resource allocation unit uses graph neural networks (GNNs) and Transformer models to perform deep modeling and sequence analysis on user data and dynamically adjust the allocation of learning resources;

[0101] The resource allocation unit further models and performs sequence analysis on user data through graph neural network (GNN) and Transformer model, and realizes resource optimization configuration by combining meta-learning and adaptive learning rate algorithm.

[0102] The specific implementation steps are as follows:

[0103] Step 1: Use multimodal learning combined with graph neural network (GNN) to model the relationship network between users, perform fusion analysis on the multimodal data of users, and generate user feature representation. The calculation formula is as follows:

[0104]

[0105] in, represents the feature representation of node v at the k+1th layer, σ is a nonlinear activation function, N(v) is the set of neighbor nodes of node v,

[0106] c vu is the normalization factor, W (k) is the learnable weight matrix of the kth layer;

[0107] Step 2: Use the Transformer model based on the self-attention mechanism to perform sequence modeling on the user's historical behavior data, generate a personalized learning path, and optimize the parameters through meta-learning and adaptive learning rate algorithm. The calculation formula is as follows:

[0108]

[0109] Among them, Q is the query matrix, K is the key matrix, and V is the value matrix. is the dimension of the key vector;

[0110] Through meta-learning and adaptive learning rate algorithms (such as Adagrad or RMSProp), the learning path is dynamically adjusted to optimize the parameter θ:

[0111]

[0112] Where Gt is the accumulated squared gradient, ∈ is the smoothing term;

[0113] Step 3: Use non-negative matrix factorization (NMF) to preliminarily model the user feature matrix and resource utility matrix, and further optimize the resource allocation strategy through diffusion models. The calculation formula is as follows:

[0114]

[0115] Among them, pθ(x t-1 ∣x t ) is the given current state x t Generate the previous state x t-1 The conditional probability, μθ(x t ,t) and is the mean and variance predicted by the neural network, and I is the identity matrix;

[0116] Step 4: Combine real-time sentiment analysis technology to dynamically adjust learning content and feedback expression. The calculation formula is as follows;

[0117]

[0118] Among them, Q(s,a) is the state-action value function, s is the current state, a is the current action, α is the learning rate, r is the immediate reward, γ is the discount factor, a′ is the possible future action, and s′ is the next state;

[0119] Step 5: Use Differential Privacy and Federated Learning methods to protect user data.

[0120] Further, such as Figure 3 As shown in the figure, the course generation module generates personalized course content based on the user's learning progress and performance through a convolutional neural network (CNN) model and a random forest algorithm;

[0121] The course generation module is responsible for generating personalized course content and providing users with dynamically adjusted learning materials based on their learning progress and performance. By analyzing users' performance data (such as learning speed, accuracy, interaction, etc.), the system can continuously optimize course content.

[0122] This module uses a convolutional neural network (CNN) model to extract features from users’ images and visual data, and combines it with a random forest algorithm to analyze users’ performance to generate personalized course content. The random forest algorithm is used in this process to evaluate users’ learning outcomes and select appropriate learning paths and materials based on different features, thereby ensuring that users can learn in the best way possible.

[0123] Specifically, the course generation module includes:

[0124] The content generation unit uses an improved convolutional neural network (CNN) and stochastic gradient descent (SGD) optimization algorithm to automatically generate diverse personalized course content based on user learning needs and interests;

[0125] Data fusion unit, which efficiently integrates multimodal learning data (text, images, videos), processes data and extracts features through deep learning technology, and generates course content that suits the user's learning style;

[0126] The dynamic adjustment unit uses deep reinforcement learning algorithms to intelligently adjust the course difficulty, teaching strategy and content presentation method based on the user's learning progress and real-time performance to adapt to changes in the user's learning status.

[0127] Further, such as Figure 4 As shown in the figure, the real-time interaction module analyzes the user's interactive content through convolutional neural network (CNN) and natural language processing (NLP) technology, and generates real-time learning feedback and personalized suggestions;

[0128] The real-time interaction module is used to analyze the user's interactive content during the learning process and generate real-time learning feedback and personalized suggestions to improve user engagement and learning outcomes.

[0129] This module uses convolutional neural network (CNN) and natural language processing (NLP) technology to analyze multimodal data input by users, such as voice, text, and gestures. The CNN model is used to process the user's image and gesture data, and NLP technology is used to parse the user's text and voice input. Through the combination of these technologies, the system can capture the user's learning status and emotional changes in real time, generate adaptive learning suggestions, and improve the effectiveness and fun of learning interactions.

[0130] Specifically, the real-time interaction module includes:

[0131] The interactive analysis unit extracts interactive text features through word embedding method based on bidirectional long short-term memory network (Bi-LSTM) and convolutional neural network (CNN);

[0132] The emotion recognition unit fuses the extracted text features with the user’s voice and visual data through a multimodal emotion fusion network;

[0133] The feedback generation unit generates personalized learning feedback based on the Transformer model and dynamically adjusts the feedback strategy through the reinforcement learning algorithm.

[0134] Further, such as Figure 5 As shown, the evaluation module uses a reinforcement learning algorithm to automatically evaluate the works submitted by users and generate feedback;

[0135] The evaluation module is responsible for automatically evaluating the works submitted by users and generating feedback to help users understand their learning progress and performance in a timely manner.

[0136] This module uses a reinforcement learning algorithm to automatically evaluate the learning works submitted by users. The reinforcement learning algorithm sets a reward mechanism based on the quality, innovation, and complexity of the user's work, and generates more accurate and personalized feedback by continuously optimizing the evaluation model. This evaluation method overcomes the subjectivity of traditional evaluation methods and provides more objective and scientific evaluation results.

[0137] Specifically, the evaluation module includes:

[0138] A multi-dimensional evaluation unit, which is used to automatically evaluate and analyze user-submitted artworks in terms of composition, color, technique, and creativity based on computer vision technology and machine learning algorithms;

[0139] The adaptive evaluation engine unit uses the user's historical performance data and real-time feedback to dynamically adjust the weight of each evaluation dimension to achieve personalized evaluation standards;

[0140] The safety optimization unit is used to optimize the evaluation criteria and strategies using the deep reinforcement learning algorithm DQN.

[0141] Furthermore, the interactive control module optimizes the learning interface using speech recognition, gesture recognition, and image recognition technologies;

[0142] The interactive control module optimizes the interactive interface between the user and the system, ensuring that the user can use the system to learn naturally and smoothly.

[0143] This module uses speech recognition, gesture recognition and image recognition technologies to create a more intuitive and user-friendly user interface. Speech recognition technology is used to recognize user voice commands, gesture recognition technology is used to detect user gesture operations, and image recognition technology is used to process learning materials and works uploaded by users, ensuring that the system can accurately understand and respond to user operations and improve user experience.

[0144] Specifically, the interaction control module includes:

[0145] The recognition unit uses a deep neural network (DNN) model to recognize the user's voice, gesture, and image input and convert the instructions into operation commands;

[0146] The interface optimization unit adjusts the layout, content presentation order and visual effects of the learning interface in real time according to the recognition results.

[0147] Furthermore, the data security module combines blockchain and differential privacy technology to protect user data and monitor data access behavior in real time.

[0148] The data security module is responsible for protecting the security and privacy of user data, while monitoring data access behavior in real time to prevent data leakage and unauthorized access.

[0149] This module combines blockchain technology and differential privacy technology to provide multiple data protection measures. Blockchain technology is used to record and monitor all data access behaviors to ensure data transparency and immutability. Differential privacy technology adds random noise during data analysis to protect users' sensitive information from being leaked. In addition, the data security module also integrates an abnormal behavior detection mechanism to monitor data access behaviors in the system in real time to prevent potential data leaks and illegal access.

[0150] Specifically, the data security module includes:

[0151] The encryption unit uses the Advanced Encryption Standard (AES-256) algorithm combined with distributed blockchain technology to encrypt and distribute user data to ensure data integrity and security;

[0152] Privacy protection unit, which protects user data based on differential privacy technology to prevent the leakage of sensitive information during data analysis;

[0153] The monitoring unit uses behavioral analysis algorithms and zero-knowledge proof methods to monitor and detect data access behavior in real time, identify abnormal access, and support multi-party collaboration based on federated learning.

[0154] Example 2

[0155] like Figure 6 As shown in FIG. 1 , an embodiment of the present invention is provided. The embodiment provides an AI-based art teaching management method, which is applied to a computing device. The method includes the following steps:

[0156] Step S1: Collecting multimodal learning data of the user, including but not limited to text, image, voice and video data;

[0157] Among them, text data is segmented and parsed using natural language processing technology; image data is collected by an image sensor with a resolution of 256x256 pixels and saved in JPEG format; the voice data sampling rate is 16kHz and stored in WAV format; the video data frame rate is 30fps and is encoded in MP4 format;

[0158] Step S2: extracting and fusing the collected multimodal data through the multimodal deep integrated learning algorithm MMDILA, wherein a convolutional neural network is used to extract features from image data, a bidirectional long short-term memory network is used to perform semantic analysis on text data, and a sentiment recognition model is used to process speech data. The Transformer model with a self-attention mechanism is used to fuse information from different data modalities to generate a comprehensive learning profile of the user;

[0159] Step S3: Based on the user learning portrait, dynamically adjust the personalized course content and teaching strategy through the reinforcement learning algorithm, wherein the reinforcement learning algorithm is based on the policy gradient method, the learning rate is 0.01, the discount factor is 0.9, and the optimization goal is to maximize the user learning effect and satisfaction.

[0160] Furthermore, the integration process of the multi-level deep learning model further includes the following steps:

[0161] The integration process of the multi-level deep learning model further includes the following steps:

[0162] Step 1: Through the multi-task learning framework, the outputs of the CNN model and the Bi-LSTM model are jointly trained in the shared layer;

[0163] Step 2: Apply self-supervised learning technology to pre-train users’ multimodal interaction data;

[0164] Step 3: Use the self-attention mechanism to dynamically adjust the importance of different data modalities.

[0165] Furthermore, the step of dynamically adjusting personalized course content and teaching strategies includes:

[0166] Step 1: Use the improved multimodal deep ensemble learning algorithm MMDILA and adversarial generative network model to predict user behavior;

[0167] Step 2: Train the generator to simulate the user's potential behavior and train the discriminator to distinguish the generated behavior from the actual behavior;

[0168] Step 3: Based on the results of the above adversarial training, adjust the user's learning path and feedback mechanism in real time.

[0169] Example 3

[0170] This embodiment realizes a more intelligent and personalized art teaching management system through an improved multimodal deep integrated learning algorithm (MMDILA) and a reinforcement learning algorithm. The present invention can dynamically analyze a user's various learning behaviors and data, generate course content that meets individual needs, and adjust learning paths and strategies in real time. At the same time, the present invention also ensures the security and privacy of user data through a variety of advanced data security technologies.

[0171] In the prior art, art teaching systems usually only use a single data source (such as text or simple learning progress data) to generate courses, lacking the ability to integrate and process multimodal data. The present invention greatly improves the depth and breadth of data fusion and analysis through an improved multimodal deep integrated learning algorithm (MMDILA).

[0172] The present invention can simultaneously process multiple types of data such as text, images, voice and video, extract image features through convolutional neural networks (CNN), perform semantic analysis through natural language processing (NLP) technology, and process voice data through emotion recognition models. Through the fusion of these multimodal data, the system can more accurately understand the user's learning habits, interests and emotional state, thereby generating more personalized learning content. Compared with the prior art, the present invention can more comprehensively capture the user's learning behavior and preferences, effectively improving the pertinence and effectiveness of personalized teaching.

[0173] Existing art teaching management systems lack real-time interactive feedback mechanisms and usually rely on fixed rules or simple scoring systems, which are unable to adjust the learning path based on users' immediate feedback. The present invention introduces a reinforcement learning algorithm to achieve real-time monitoring and dynamic optimization of user learning behavior.

[0174] The present invention uses a reinforcement learning algorithm to evaluate the user's learning effect and behavior in real time, and the system continuously adjusts itself to improve the learning effect. For example, when a user shows difficulty in a certain knowledge point, the system will automatically increase the learning resources of that part of the content or adjust the teaching strategy. Through this dynamic optimization mechanism, the system can generate a personalized path that better meets the user's learning needs based on the user's feedback and performance.

[0175] The real-time feedback capability and dynamic optimization mechanism of the present invention significantly improve teaching effects and user satisfaction, greatly reduce learning time, and increase user learning involvement compared to the prior art.

[0176] In the prior art, the user data security measures of the art teaching system are relatively basic, mainly relying on simple encryption or access control. On this basis, the present invention combines blockchain technology and differential privacy technology to greatly enhance data security and privacy protection capabilities. The present invention uses blockchain technology to record and monitor data access behavior to ensure the transparency and non-tamperability of data operations. At the same time, differential privacy technology is used to protect users' sensitive data and prevent personal privacy from being leaked during data analysis and use. In addition, the system integrates a behavioral anomaly detection model that can monitor and detect potential data leaks and illegal access in real time. Compared with the prior art, the present invention not only provides higher security during data storage and transmission, but also has made breakthrough progress in privacy protection and data anomaly monitoring, making user data more secure and reliable.

[0177] This embodiment significantly improves the personalized teaching ability, real-time feedback mechanism and data security of the existing art teaching management system by introducing multimodal deep integration learning algorithms, reinforcement learning algorithms and multiple data security technologies. Through these innovative technical improvements, the present invention achieves comprehensive improvements in user experience, learning effects and data protection.

[0178] Example 4

[0179] In order to verify the significant advantages of the present invention in personalized teaching and data security, two groups of experiments were conducted: one is to compare the personalized teaching effects of the system of the present invention with that of the traditional art teaching system, and the other is to verify the performance of the data security module in preventing data leakage and protecting privacy.

[0180] Experimental design

[0181] Experimental sample: 100 participants were selected and divided into an experimental group and a control group, with 50 people in each group. The experimental group used the AI ​​art teaching management system of the present invention, and the control group used the traditional art teaching management system.

[0182] Experimental period: 4 weeks.

[0183] Test indicators: learning effect score, learning time, user satisfaction, data leakage detection success rate, and privacy protection rate.

[0184] Experimental Results

[0185] index Experimental group (system of the present invention) Control group (traditional system) Improvement rate Learning effect score (out of 100 points) 85.2 63.1 35.00％ Average study time (hours) 28.6 35.8 -20.10％ User satisfaction (out of 5 points) 4.7 3.5 34.30％ Data leakage detection success rate (times) 20 / 20 December 20 66.70％ Privacy protection rate (percentage) 98.50％ 85.00％ 15.90％

[0186] Experimental data shows that the learning effect score: during the 4-week learning cycle, the average learning effect score of the experimental group was 85.2 points, which was significantly higher than the 63.1 points of the control group, indicating that the system of the present invention significantly improves the personalized teaching effect.

[0187] Average learning time: The average learning time of the experimental group was 28.6 hours, which was 20.1% less than the 35.8 hours of the control group. This shows that through real-time feedback and personalized path optimization, the system of the present invention effectively reduces the time required for learning.

[0188] User satisfaction: The user satisfaction score of the experimental group using the system of the present invention was 4.7 points, which was higher than the 3.5 points of the control group. The results show that the system of the present invention provides a better user experience and learning participation.

[0189] Data leakage detection success rate: In 20 simulated attack tests, the system of the present invention successfully detected all 20 data leakage attempts, with a success rate of 100%, while the control group had a success rate of 60% (12 / 20). The security protection capability of the system of the present invention is stronger.

[0190] Privacy protection rate: The privacy protection rate of the system of the present invention is 98.5%, which is significantly better than 85.0% of the control group, indicating that the present invention performs well in protecting user data privacy.

[0191] Experimental data show that the AI ​​art teaching management system of the present invention is significantly superior to the existing technology in terms of personalized teaching effects, optimization of learning time, improvement of user satisfaction, and data security and privacy protection.

[0192] In summary, the present invention innovatively uses an improved multimodal deep integrated learning algorithm (MMDILA) to significantly improve data fusion and analysis capabilities. Unlike the prior art, the present invention can simultaneously process multiple data types of users (such as text, images, voice, and video), and efficiently analyze these data through convolutional neural networks (CNN), natural language processing (NLP), and emotion recognition models. This innovative data processing method achieves a more comprehensive and accurate understanding of user behavior, enabling the present invention to generate and dynamically adjust learning paths according to the personalized needs of users, significantly improving the effectiveness and pertinence of personalized teaching.

[0193] The present invention introduces a reinforcement learning algorithm and establishes a real-time feedback and dynamic optimization mechanism, which can continuously adjust the course content and teaching strategy according to the user's real-time performance and feedback. Compared with the existing technology that relies on fixed rules or simple scoring feedback, the present invention can provide more immediate and accurate personalized suggestions, thereby effectively shortening the learning time and improving learning efficiency. This innovative interactive mechanism enhances the user experience and increases the user's learning engagement and satisfaction.

[0194] The present invention proposes a new data security and privacy protection solution by integrating blockchain technology and differential privacy technology. Blockchain technology ensures the transparency and immutability of data operations, while differential privacy technology effectively prevents the leakage of user sensitive information during data analysis and use. In addition, the present invention also introduces an anomaly detection model that can monitor and respond to potential data leakage and illegal access risks in real time. Compared with the single encryption or access control measures in the prior art, the present invention achieves a higher level of protection in data security and privacy protection.

[0195] The present invention adopts a modular system architecture design, and each functional module (such as user management, course generation, interactive feedback, data security, etc.) is independent of each other but works in coordination. This architectural design not only improves the scalability and compatibility of the system, facilitates subsequent function development and system upgrades, but also reduces the maintenance cost and complexity of the system. Compared with the relatively closed and inflexible system architecture in the prior art, the design of the present invention can better adapt to different user needs and usage scenarios, and has a wide range of market application potential.

[0196] The present invention has verified its significant advantages in personalized teaching effect, learning time optimization, user satisfaction improvement, and data security and privacy protection through a large amount of experimental data. The experimental results show that the personalized teaching effect score of the system of the present invention is improved by 35%, the learning time is reduced by 20%, the user satisfaction is increased by 34%, and it performs well in data leakage detection and privacy protection.

[0197] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0198] Any process or method description in the flow chart or otherwise described herein can be understood to represent a module, fragment or portion of a code including one or more executable instructions for implementing the steps of a specific logical function or process. And the scope of the preferred embodiment of the present application includes other implementations, in which the functions may not be performed in the order shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved.

[0199] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An AI-based art teaching management system, characterized in that: The system comprises: An intelligent user management module is used to identify the user identity and dynamically adjust the allocation of learning resources and permission configuration using a support vector machine (SVM) algorithm based on the user's learning behavior data and historical records. The intelligent user management module is based on an improved multimodal deep integration learning algorithm MMDILA, which optimizes the data preprocessing and analysis process by performing unified modeling, feature selection, and data fusion on multimodal data; The course generation module generates personalized course content based on the user's learning progress and performance through the convolutional neural network (CNN) model and random forest algorithm; The real-time interaction module uses convolutional neural networks (CNN) and natural language processing (NLP) to analyze user interaction content and generate real-time learning feedback and personalized suggestions. The evaluation module uses reinforcement learning algorithms to automatically evaluate user-submitted works and generate feedback; Interactive control module, which uses speech recognition, gesture recognition and image recognition technologies to optimize the learning interface; The data security module combines blockchain and differential privacy technology to protect user data and monitor data access behavior in real time.

2. The AI-based art teaching management system according to claim 1 is characterized in that: The intelligent user management module further includes: The user identification unit uses a deep neural network AI algorithm to efficiently and accurately identify users and provide personalized access rights control; The data analysis unit uses multimodal learning data, including text, images, and voice, for comprehensive analysis to generate personalized learning paths and progress plans that meet user characteristics; The resource allocation unit uses graph neural networks and Transformer models to perform deep modeling and sequence analysis on user data and dynamically adjust the allocation of learning resources; The resource allocation unit further models and performs sequence analysis on user data through graph neural networks and Transformer models, and combines meta-learning and adaptive learning rate algorithms to achieve resource optimization configuration. The specific implementation steps are as follows: Step 1: Use multimodal learning combined with graph neural networks to model the relationship network between users, perform fusion analysis on the multimodal data of users, and generate user feature representation. The calculation formula is as follows: in, represents the feature representation of node v at the k+1th layer, σ is a nonlinear activation function, N(v) is the set of neighbor nodes of node v, c vu is the normalization factor, W (k) is the learnable weight matrix of the kth layer; Step 2: Use the Transformer model based on the self-attention mechanism to perform sequence modeling on the user's historical behavior data, generate a personalized learning path, and optimize the parameters through meta-learning and adaptive learning rate algorithm. The calculation formula is as follows: Among them, Q is the query matrix, K is the key matrix, and V is the value matrix. is the dimension of the key vector; Through meta-learning and adaptive learning rate algorithm, the learning path is dynamically adjusted to optimize the parameter θ: Where Gt is the accumulated squared gradient, ∈ is the smoothing term; Step 3: Use non-negative matrix factorization to preliminarily model the user feature matrix and resource utility matrix, and further optimize the resource allocation strategy through the diffusion model. The calculation formula is as follows: Among them, pθ(x t-1 ∣x t ) is the given current state x t Generate the previous state x t-1 The conditional probability, μθ(x t ,t) and is the mean and variance predicted by the neural network, and I is the identity matrix; Step 4: Combine real-time sentiment analysis technology to dynamically adjust learning content and feedback expression. The calculation formula is as follows; Among them, Q(s,a) is the state-action value function, s is the current state, a is the current action, α is the learning rate, r is the immediate reward, γ is the discount factor, a′ is the possible future action, and s′ is the next state; Step 5: Use differential privacy and federated learning methods to protect user data.

3. The AI-based art teaching management system according to claim 1 is characterized in that: The course generation module includes: The content generation unit uses convolutional neural networks and stochastic gradient descent optimization algorithms to automatically generate diverse personalized course content based on user learning needs and interests; Data fusion unit, which efficiently integrates multimodal learning data, processes data and extracts features through deep learning technology, and generates course content that suits the user's learning style; The dynamic adjustment unit uses deep reinforcement learning algorithms to intelligently adjust the course difficulty, teaching strategy and content presentation method based on the user's learning progress and real-time performance.

4. The AI-based art teaching management system according to claim 1 is characterized in that: The real-time interactive module includes: The interaction analysis unit extracts interactive text features through word embedding method based on bidirectional long short-term memory network and convolutional neural network; the emotion recognition unit fuses the extracted text features with the user's voice and visual data through a multimodal emotion fusion network; the feedback generation unit generates personalized learning feedback based on the Transformer model, and dynamically adjusts the feedback strategy through a reinforcement learning algorithm.

5. The AI-based art teaching management system according to claim 1 is characterized in that: The evaluation module includes: a multi-dimensional evaluation unit for automatically evaluating and analyzing the artworks submitted by users from multiple aspects such as composition, color, technique and creativity based on computer vision technology and machine learning algorithms; The adaptive evaluation engine unit uses the user's historical performance data and real-time feedback to dynamically adjust the weight of each evaluation dimension to achieve personalized evaluation standards; The safety optimization unit is used to optimize the evaluation criteria and strategies using the deep reinforcement learning algorithm DQN.

6. The AI-based art teaching management system according to claim 1 is characterized in that: The interactive control module comprises: The recognition unit uses a deep neural network model to recognize the user's voice, gesture, and image input and converts the instructions into operation commands; The interface optimization unit adjusts the layout, content presentation order and visual effects of the learning interface in real time according to the recognition results.

7. The AI-based art teaching management system according to claim 1 is characterized in that: The data security module comprises: The encryption unit uses the Advanced Encryption Standard AES-256 algorithm combined with distributed blockchain technology to encrypt and distribute user data; The privacy protection unit protects user data based on differential privacy technology to prevent the leakage of sensitive information during data analysis. The monitoring unit uses behavioral analysis algorithms and zero-knowledge proof methods to monitor and detect data access behavior in real time, identify abnormal access, and support multi-party collaboration based on federated learning.

8. An AI-based art teaching management method according to any one of claims 1 to 7, applied to a computing device, the method comprising the following steps: Step S1: Collecting multimodal learning data of the user, including but not limited to text, image, voice and video data; Among them, text data is segmented and parsed using natural language processing technology; image data is collected by an image sensor with a resolution of 256x256 pixels and saved in JPEG format; the voice data sampling rate is 16kHz and stored in WAV format; the video data frame rate is 30fps and is encoded in MP4 format; Step S2: extracting and fusing the collected multimodal data through the multimodal deep integrated learning algorithm MMDILA, wherein a convolutional neural network is used to extract features from image data, a bidirectional long short-term memory network is used to perform semantic analysis on text data, and a sentiment recognition model is used to process speech data. The Transformer model with a self-attention mechanism is used to fuse information from different data modalities to generate a comprehensive learning profile of the user; Step S3: Based on the user learning portrait, dynamically adjust the personalized course content and teaching strategy through the reinforcement learning algorithm, wherein the reinforcement learning algorithm is based on the policy gradient method, the learning rate is 0.01, the discount factor is 0.9, and the optimization goal is to maximize the user learning effect and satisfaction.

9. The AI-based art teaching management method according to claim 8 is characterized in that: The integration process of the multi-level deep learning model further includes the following steps: Step 1: Through the multi-task learning framework, the outputs of the CNN model and the Bi-LSTM model are jointly trained in the shared layer; Step 2: Apply self-supervised learning technology to pre-train users’ multimodal interaction data; Step 3: Use the self-attention mechanism to dynamically adjust the importance of different data modalities.

10. The AI-based art teaching management method according to claim 8, characterized in that: The steps of dynamically adjusting personalized course content and teaching strategies include: Step 1: Use the improved multimodal deep ensemble learning algorithm MMDILA and adversarial generative network model to predict user behavior; Step 2: Train the generator to simulate the user's potential behavior and train the discriminator to distinguish the generated behavior from the actual behavior; Step 3: Based on the results of the above adversarial training, adjust the user's learning path and feedback mechanism in real time.

Images