Traditional culture display system based on artificial intelligence

Abstract

The invention discloses a traditional culture display system based on artificial intelligence. The traditional culture display system comprises a front-end display layer, a middle-end interaction layer and a rear-end data processing layer, the rear-end data processing layer comprises a big data platform module, a data generation and enhancement module, a cost-sensitive learning module and a machine learning algorithm module; the data generation and enhancement module obtains original data, generates analog data by using a GAN, fuses the analog data with real data and performs enhancement processing on the fused data; the cost-sensitive learning module and the machine learning algorithm module collect enhanced data provided by the data generation and enhancement module, and the cost-sensitive learning module balances the influence of different types of samples in the enhanced data in model training according to the cost distribution weights of the different types of samples. And the machine learning algorithm module adjusts a loss function or an optimization target in a training process according to weight information provided by the cost-sensitive learning module.

Description

Technical Field

[0001] The present invention relates to the technical field of traditional culture display, and in particular to a traditional culture display system based on artificial intelligence. Background Art

[0002] With the rapid development of artificial intelligence technology, its application in various fields is becoming more and more extensive. Artificial intelligence can realize the intelligent processing and transformation of traditional cultural content, so that it can be presented to the audience in a more vivid and intuitive form. As a treasure of national spirit, traditional culture is of great significance to enhancing the country's cultural soft power and strengthening national cohesion. However, with the rapid development of modern society, the inheritance and development of traditional culture faces many challenges. Therefore, using artificial intelligence technology to display and inherit traditional culture has become a new trend and demand.

[0003] The data set of the traditional culture display system may be relatively sparse, that is, there is less interaction data between users and traditional cultural content. This may make it difficult for the model to capture the user's real interest points during the training process, and thus it is easy to overfit to the limited training data. Then the traditional cultural scenes and interactive content seen by users on the VR / AR interface or touch screen may become monotonous and boring. This monotony may reduce the user's overall experience of the system, making them feel a lack of exploration fun and surprise of discovery. Therefore, a traditional culture display system based on artificial intelligence is proposed. Summary of the invention

[0004] The purpose of this invention is to solve the problem in the prior art that data collection will collect a large amount of multidimensional data, including vehicle flow, speed, type, weather conditions, social media reports, etc. When the dimensions of these data are very high, it will not only increase the computational complexity, but also may lead to "dimensionality disaster", that is, as the dimension increases, the sparsity and noise of the data will also increase, thereby affecting the accuracy and efficiency of the model. A traditional cultural display system based on artificial intelligence is proposed.

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

[0006] A traditional cultural display system based on artificial intelligence, including a front-end display layer, a mid-end interaction layer, and a back-end data processing layer;

[0007] The front-end display layer includes a VR / AR interface module and a touch screen and interactive device module. The VR / AR interface module and the touch screen and interactive device module obtain 3D models, animations, sound effects and other resources in the traditional cultural database from the big data platform module. The VR / AR interface module receives voice instructions or text information provided by the natural language processing module and converts them into corresponding actions or feedback in the scene. The VR / AR interface module transmits the user's emotional state information to the emotion recognition and feedback module. The touch screen and interactive device module receives recommended content or instructions provided by the natural language processing module or the personalized recommendation module.

[0008] The middle-end interaction layer includes a natural language processing module, an emotion recognition and feedback module, and a personalized recommendation module. The emotion recognition and feedback module transmits the user's emotional state information back to the data processing layer. The personalized recommendation module obtains the user's interests and historical behaviors from the big data platform module, and builds user portraits and content tags based on the personalized model provided by the machine learning algorithm module.

[0009] The back-end data processing layer includes a big data platform module, a data generation and enhancement module, a cost-sensitive learning module, and a machine learning algorithm module: the big data platform module provides the machine learning algorithm module with data resources in the traditional cultural database, the data generation and enhancement module obtains detailed records of the interaction between users and traditional cultural content in the traditional cultural database and the corresponding traditional cultural content information from the big data platform module and uses GAN to generate simulated data, fuses the simulated data with the real data and enhances the fused data, the cost-sensitive learning module and the machine learning algorithm module collect the enhanced data provided by the data generation and enhancement module, and the cost-sensitive learning module uses the cost-sensitive learning method to balance the influence of different categories of samples in the enhanced data in the model training. High weights are given to low-frequency content, and the machine learning algorithm module adjusts the loss function or optimization target according to the weight information provided by the cost-sensitive learning module during the training process.

[0010] The above technical solution further includes:

[0011] Furthermore, the VR / AR interface module includes a rendering engine unit, an interaction processing unit, an emotion recognition unit and a data communication unit. The rendering engine unit is responsible for rendering 3D models, animations and sound effects into a virtual environment visible to the user, obtaining resources from the big data platform module, and updating elements in the virtual environment in real time. The interaction processing unit receives the user's voice commands or text information provided by the natural language processing module, and converts them into system-recognizable commands. The emotion recognition unit identifies the user's emotional state by analyzing the user's facial expressions and emotional information of voice intonation, and transmits the emotional state information to the emotion recognition and feedback module. The data communication unit is responsible for data communication between the VR / AR interface module and other modules.

[0012] Furthermore, the data generation and enhancement module includes a data acquisition unit, a GAN model unit, a data fusion and enhancement unit, and a data evaluation and feedback unit. The data acquisition unit is responsible for acquiring detailed records of user interactions with traditional cultural content and corresponding traditional cultural content information from the big data platform module, and preprocessing the acquired data, such as cleaning, denoising, format conversion, etc., to ensure the quality and availability of the data. The GAN model unit uses the GAN model to generate simulated data similar to real data. The data fusion and enhancement unit fuses the generated simulated data with the real data, and performs enhancement processing on the fused data, such as data amplification, data transformation, etc., to improve the diversity and generalization ability of the data. The data evaluation and feedback unit performs quality assessment on the generated simulated data and enhanced data to ensure that they can meet the needs of the system.

[0013] Furthermore, the specific steps of the GAN model unit using GAN to generate new virtual data are:

[0014] Data collection: Collect the existing processed interaction data between users and traditional cultural content;

[0015] Feature extraction: extract key features related to user interests, preferences, and traditional cultural content characteristics from the preprocessed interaction data;

[0016] GAN model training: Build a GAN model, including a generator and a discriminator. The task of the generator is to generate virtual data similar to real data, while the task of the discriminator is to distinguish between real data and virtual data. Through training, the generator can generate virtual data that is closer and closer to real data.

[0017] Simulated data generation: Use the trained generator to generate new simulated data.

[0018] Furthermore, in feature extraction, PCA is used to extract key variables that can represent user behavior and traditional cultural content characteristics. The specific steps are:

[0019] Data standardization: Before performing PCA, the data is standardized by subtracting the mean and dividing by the standard deviation so that the mean of each feature is 0 and the variance is 1. The calculation formula is Among them, X is the original data, μ is the mean, σ is the standard deviation, and Z is the standardized data;

[0020] Calculate the covariance matrix: The covariance matrix is ​​a matrix that describes the relationship between the various features in the data set. Its elements represent the covariance between different features. The covariance matrix is ​​expressed as Where n is the number of samples, Z T is the transposed matrix of Z;

[0021] Solve the eigenvalue and eigenvector: Perform eigendecomposition on the covariance matrix C to obtain the eigenvalue λ and the corresponding eigenvector v. The eigenvalue λ represents the variance explained by each principal component, and the eigenvector v represents the direction of each principal component in the original feature space.

[0022] Select principal components: According to the size of the eigenvalue, select the eigenvectors corresponding to the first k largest eigenvalues. The eigenvectors constitute a new feature space. Usually, the choice of k can be determined according to the cumulative contribution rate of the eigenvalue (that is, the ratio of the sum of the first k eigenvalues ​​to the sum of the total eigenvalues).

[0023] Furthermore, the specific steps of the GAN model training are:

[0024] Constructing the GAN model: Generator: Design a neural network structure whose input is a random noise vector and whose output is virtual data similar to the detailed records of the user's interaction with traditional cultural content and the corresponding traditional cultural content information; Discriminator: Another neural network structure whose input is the detailed records of the user's interaction with traditional cultural content and the real data of the corresponding traditional cultural content information or the virtual data generated by the generator, and whose output is the probability that the data is real data;

[0025] Define the loss function: Generator loss: measures the difference between the virtual data generated by the generator and the real data. The generator loss function is Where z represents a random noise vector, G(z) represents virtual data generated by the generator, and D(G(z)) represents the predicted probability of the discriminator for the virtual data generated by the generator; Discriminator loss: measures the ability of the discriminator to distinguish between real data and virtual data. The discriminator loss function is Among them, x represents the real data, and D(x) represents the predicted probability of the discriminator for the real data;

[0026] Training process: Initialize model parameters: assign random weights to the generator and discriminator; iterative training: train the discriminator: use real data and virtual data generated by the generator to train the discriminator so that it can accurately distinguish between the two; train the generator: use the feedback from the discriminator to train the generator so that the virtual data it generates is closer and closer to the real data; update weights: after each iteration, update the weights of the generator and discriminator according to the gradient of the loss function; through the iterative training process, continuously optimize the weights of the generator and discriminator so that the generator generates virtual data that is closer and closer to the real data.

[0027] Furthermore, the cost-sensitive learning module adopts a cost-sensitive learning method to balance the influence of different categories of samples in the enhanced data in model training, including the following steps:

[0028] Data preprocessing: Clean, organize and format the user interaction data of the traditional cultural display system to ensure the quality and consistency of the data, and identify and handle imbalances in the data, such as the small amount of interaction data for some traditional cultural content;

[0029] Weight allocation: weights are allocated according to the cost (or importance) of samples of different categories. For example, low-frequency traditional cultural content can be given a higher weight so that it receives more attention during model training. Weight allocation can be based on a comprehensive consideration of multiple factors such as data frequency, user interest, and content importance.

[0030] Application of cost-sensitive learning algorithm: Introduce weights in the loss function to reflect the cost differences of samples of different categories, and use weighted cross entropy loss function: Among them, w i is the weight of sample i, y i is the true label of sample i, is the predicted label of sample i, and N is the total number of samples;

[0031] Weight information transmission: transmit weighted data to the machine learning algorithm module.

[0032] Furthermore, the machine learning algorithm module includes a data preprocessing unit, a feature selection and dimensionality reduction unit, a model training unit, a model evaluation and optimization unit, and a model deployment and update unit. The data preprocessing unit preprocesses the data obtained from the big data platform module and the data generation and enhancement module. The feature selection and dimensionality reduction unit selects the most valuable features for model training from the preprocessed data and performs dimensionality reduction processing. The model training unit trains the recommendation model based on the selected feature data and the weight information provided by the cost-sensitive learning module. The model evaluation and optimization unit evaluates the trained model and optimizes and adjusts the model according to the evaluation results. The model deployment and update unit deploys the optimized model to the personalized recommendation module for users to make real-time recommendations and interactions.

[0033] The present invention has the following beneficial effects:

[0034] In the present invention, the data generation and enhancement module generates simulated data through GAN and fuses and enhances it with real data, which effectively alleviates the problem of data sparsity and improves the diversity and richness of data. The cost-sensitive learning module assigns weights to the costs of samples of different categories, especially giving high weights to low-frequency content, effectively balancing the impact of samples of different categories in the data on model training. This balancing process helps the model capture the user's real interests more comprehensively during the training process, reduces the risk of overfitting to limited training data, and improves the generalization ability of the model and the diversity of recommendation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a system block diagram of a traditional cultural display system based on artificial intelligence proposed by the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] See also Figure 1 As shown, the present invention is a traditional culture display system based on artificial intelligence, including a front-end display layer, a mid-end interaction layer and a back-end data processing layer;

[0038] The front-end display layer includes a VR / AR interface module and a touch screen and interactive device module. The VR / AR interface module and the touch screen and interactive device module obtain 3D models, animations, sound effects and other resources in the traditional cultural database from the big data platform module. The VR / AR interface module receives voice commands or text information provided by the natural language processing module and converts them into corresponding actions or feedback in the scene. The VR / AR interface module transmits the user's emotional state information to the emotion recognition and feedback module. The touch screen and interactive device module receives recommended content or instructions provided by the natural language processing module or the personalized recommendation module.

[0039] The middle-end interaction layer includes the natural language processing module, the emotion recognition and feedback module, and the personalized recommendation module. The emotion recognition and feedback module transmits the user's emotional state information back to the data processing layer. The personalized recommendation module obtains the user's interests and historical behaviors from the big data platform module, and builds user portraits and content tags based on the personalized model provided by the machine learning algorithm module.

[0040] The back-end data processing layer includes a big data platform module, a data generation and enhancement module, a cost-sensitive learning module, and a machine learning algorithm module: the big data platform module provides the machine learning algorithm module with data resources in the traditional cultural database. The data generation and enhancement module obtains detailed records of the interaction between users and traditional cultural content in the traditional cultural database and the corresponding traditional cultural content information from the big data platform module and uses GAN to generate simulated data, fuses the simulated data with the real data, and enhances the fused data. The cost-sensitive learning module and the machine learning algorithm module collect the enhanced data provided by the data generation and enhancement module. The cost-sensitive learning module uses the cost-sensitive learning method to balance the influence of different categories of samples in the enhanced data in model training. High weights are given to low-frequency content. During the training process, the machine learning algorithm module adjusts the loss function or optimization target according to the weight information provided by the cost-sensitive learning module.

[0041] In one embodiment, for the above-mentioned VR / AR interface module, the VR / AR interface module includes a rendering engine unit, an interaction processing unit, an emotion recognition unit and a data communication unit. The rendering engine unit is responsible for rendering 3D models, animations and sound effects into a virtual environment visible to the user, obtaining resources from the big data platform module, and updating the elements in the virtual environment in real time. The interaction processing unit receives the user's voice commands or text information provided by the natural language processing module, and converts them into system-recognizable commands. The emotion recognition unit identifies the user's emotional state by analyzing the user's facial expressions and emotional information of voice intonation, and transmits the emotional state information to the emotion recognition and feedback module. The data communication unit is responsible for data communication between the VR / AR interface module and other modules.

[0042] In one embodiment, for the above-mentioned data generation and enhancement module, the data generation and enhancement module includes a data acquisition unit, a GAN model unit, a data fusion and enhancement unit, and a data evaluation and feedback unit. The data acquisition unit is responsible for obtaining detailed records of user interactions with traditional cultural content and corresponding traditional cultural content information from the big data platform module, and preprocessing the acquired data, such as cleaning, denoising, format conversion, etc., to ensure the quality and availability of the data. The GAN model unit uses the GAN model to generate simulated data similar to real data. The data fusion and enhancement unit fuses the generated simulated data with the real data, and performs enhancement processing on the fused data, such as data amplification, data transformation, etc., to improve the diversity and generalization ability of the data. The data evaluation and feedback unit performs quality evaluation on the generated simulated data and enhanced data to ensure that they can meet the needs of the system.

[0043] In one embodiment, for the above-mentioned GAN model unit, the specific steps of the GAN model unit using GAN to generate new virtual data are:

[0044] Data collection: Collect the existing processed interaction data between users and traditional cultural content;

[0045] Feature extraction: extract key features related to user interests, preferences, and traditional cultural content characteristics from the preprocessed interaction data;

[0046] GAN model training: Build a GAN model, including a generator and a discriminator. The task of the generator is to generate virtual data similar to real data, while the task of the discriminator is to distinguish between real data and virtual data. Through training, the generator can generate virtual data that is closer and closer to real data.

[0047] Simulated data generation: Use the trained generator to generate new simulated data.

[0048] In one embodiment, for the above feature extraction, PCA is used to extract key variables that can represent user behavior and traditional cultural content characteristics. The specific steps are:

[0049] Data standardization: Before performing PCA, the data is standardized by subtracting the mean and dividing by the standard deviation so that the mean of each feature is 0 and the variance is 1. The calculation formula is Among them, X is the original data, μ is the mean, σ is the standard deviation, and Z is the standardized data;

[0050] Calculate the covariance matrix: The covariance matrix is ​​a matrix that describes the relationship between the various features in the data set. Its elements represent the covariance between different features. The covariance matrix is ​​expressed as Where n is the number of samples, Z T is the transposed matrix of Z;

[0051] Solve the eigenvalue and eigenvector: Perform eigendecomposition on the covariance matrix C to obtain the eigenvalue λ and the corresponding eigenvector v. The eigenvalue λ represents the variance explained by each principal component, and the eigenvector v represents the direction of each principal component in the original feature space.

[0052] Select principal components: According to the size of the eigenvalue, select the eigenvectors corresponding to the first k largest eigenvalues. The eigenvectors constitute a new feature space. Usually, the choice of k can be determined according to the cumulative contribution rate of the eigenvalue (that is, the ratio of the sum of the first k eigenvalues ​​to the sum of the total eigenvalues).

[0053] Construct the fused feature vector: Project the original feature matrix X onto the new feature space to obtain the fused feature vector. Specifically, for each sample, perform matrix multiplication between its original feature vector and the selected feature vector to obtain the representation of the sample in the new feature space (i.e., the fused feature vector).

[0054] Example

[0055] Suppose we have a two-dimensional data set containing 5 samples, each with two features. The data is as follows:

[0056]

[0057] Data standardization: Calculate the mean and standard deviation, and then perform standardization.

[0058] Calculate the covariance matrix: Get the covariance matrix:

[0059]

[0060] Solve the eigenvalues ​​and eigenvectors: Calculate the eigenvalues: λ1 = 1.28402771, λ2 = 0.04908304, and the corresponding eigenvectors;

[0061] Select the principal components:

[0062] Since λ1 is much larger than λ2, we select the eigenvector corresponding to λ1 as the principal component.

[0063] Convert data to the new space:

[0064] Project the original data onto the principal components to obtain the reduced-dimensional data.

[0065] In one embodiment, for the above GAN model training, the specific steps of GAN model training are:

[0066] Constructing the GAN model: Generator: Design a neural network structure whose input is a random noise vector and whose output is virtual data similar to the detailed records of the user's interaction with traditional cultural content and the corresponding traditional cultural content information; Discriminator: Another neural network structure whose input is the detailed records of the user's interaction with traditional cultural content and the real data of the corresponding traditional cultural content information or the virtual data generated by the generator, and whose output is the probability that the data is real data;

[0067] Define the loss function: Generator loss: measures the difference between the virtual data generated by the generator and the real data. The generator loss function is Where z represents a random noise vector, G(z) represents virtual data generated by the generator, and D(G(z)) represents the predicted probability of the discriminator for the virtual data generated by the generator; Discriminator loss: measures the ability of the discriminator to distinguish between real data and virtual data. The discriminator loss function is Among them, x represents the real data, and D(x) represents the predicted probability of the discriminator for the real data;

[0068] Training process: Initialize model parameters: assign random weights to the generator and discriminator; iterative training: train the discriminator: use real data and virtual data generated by the generator to train the discriminator so that it can accurately distinguish between the two; train the generator: use the feedback from the discriminator to train the generator so that the virtual data it generates is closer and closer to the real data; update weights: after each iteration, update the weights of the generator and discriminator according to the gradient of the loss function; through the iterative training process, continuously optimize the weights of the generator and discriminator so that the generator generates virtual data that is closer and closer to the real data.

[0069] In one embodiment, for the above-mentioned cost-sensitive learning module, the cost-sensitive learning module adopts a cost-sensitive learning method to balance the influence of different categories of samples in the enhanced data in model training, including the following steps:

[0070] Data preprocessing: Clean, organize and format the user interaction data of the traditional cultural display system to ensure the quality and consistency of the data, and identify and handle imbalances in the data, such as the small amount of interaction data for some traditional cultural content;

[0071] Weight allocation: weights are allocated according to the cost (or importance) of samples of different categories. For example, low-frequency traditional cultural content can be given a higher weight so that it receives more attention during model training. Weight allocation can be based on a comprehensive consideration of multiple factors such as data frequency, user interest, and content importance.

[0072] Application of cost-sensitive learning algorithm: Introduce weights in the loss function to reflect the cost differences of samples of different categories, and use weighted cross entropy loss function: Among them, w i is the weight of sample i, y i is the true label of sample i, is the predicted label of sample i, and N is the total number of samples;

[0073] Weight information transmission: transmit weighted data to the machine learning algorithm module.

[0074] Assume that there are three kinds of traditional cultural content in the traditional cultural display system: calligraphy, Chinese painting, and paper-cutting. Among them, paper-cutting has less user interaction data and is a low-frequency content. In order to improve the model's ability to recognize paper-cutting content, a cost-sensitive learning algorithm can be used.

[0075] Data preprocessing: Clean and organize user interaction data of calligraphy, Chinese painting, and paper-cutting.

[0076] Weight allocation: Assign weights w1, w2, and w3 to calligraphy, Chinese painting, and paper-cutting, respectively. Since paper-cutting is a low-frequency content, it can be given a higher weight, such as w3>w1, and w3>w2.

[0077] Application of cost-sensitive learning algorithm: Select weighted cross entropy loss function as the loss function and use weighted data to train the model.

[0078] In one embodiment, for the above-mentioned machine learning algorithm module, the machine learning algorithm module includes a data preprocessing unit, a feature selection and dimensionality reduction unit, a model training unit, a model evaluation and optimization unit, and a model deployment and update unit. The data preprocessing unit preprocesses the data obtained from the big data platform module and the data generation and enhancement module. The feature selection and dimensionality reduction unit selects the most valuable features for model training from the preprocessed data and performs dimensionality reduction processing. The model training unit trains the recommendation model based on the selected feature data and the weight information provided by the cost-sensitive learning module. The model evaluation and optimization unit evaluates the trained model and optimizes and adjusts the model according to the evaluation results. The model deployment and update unit deploys the optimized model to the personalized recommendation module for users to make real-time recommendations and interactions.

[0079] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A traditional culture display system based on artificial intelligence, characterized in that: It includes the front-end display layer, the middle-end interaction layer and the back-end data processing layer; The front-end display layer includes a VR / AR interface module and a touch screen and interactive device module, the VR / AR interface module and the touch screen and interactive device module obtain resources in the traditional cultural database from the big data platform module, the VR / AR interface module receives voice instructions or text information provided by the natural language processing module, and converts it into corresponding actions or feedback in the scene, the VR / AR interface module transmits the user's emotional state information to the emotion recognition and feedback module, and the VR / AR interface module and the touch screen and interactive device module receive recommended content or instructions provided by the natural language processing module or the personalized recommendation module; The middle-end interaction layer includes a natural language processing module, an emotion recognition and feedback module, and a personalized recommendation module. The emotion recognition and feedback module transmits the user's emotional state information back to the back-end data processing layer. The personalized recommendation module obtains the user's interests and historical behaviors from the big data platform module, and builds user portraits and content tags based on the personalized model provided by the machine learning algorithm module. The back-end data processing layer includes a big data platform module, a data generation and enhancement module, a cost-sensitive learning module and a machine learning algorithm module: the big data platform module provides the machine learning algorithm module with data resources in the traditional cultural database, the data generation and enhancement module obtains detailed records of the interaction between users and traditional cultural content in the traditional cultural database and the corresponding traditional cultural content information from the big data platform module and uses GAN to generate simulated data, fuses the simulated data with real data and enhances the fused data, the cost-sensitive learning module and the machine learning algorithm module collect the enhanced data provided by the data generation and enhancement module, the cost-sensitive learning module balances the influence of different categories of samples in the enhanced data in model training according to the cost allocation weights of different categories of samples, and gives high weights to low-frequency content, and the machine learning algorithm module adjusts the loss function or optimization target according to the weight information provided by the cost-sensitive learning module during the training process.

2. According to the artificial intelligence-based traditional culture display system of claim 1, it is characterized in that: The VR / AR interface module includes a rendering engine unit, an interaction processing unit, an emotion recognition unit and a data communication unit. The rendering engine unit is responsible for rendering 3D models, animations and sound effects into a virtual environment visible to the user, obtaining resources from the big data platform module, and updating elements in the virtual environment in real time. The interaction processing unit receives the user's voice instructions or text information provided by the natural language processing module, and converts it into instructions recognizable by the system. The emotion recognition unit identifies the user's emotional state by analyzing the user's facial expressions and emotional information of voice intonation, and transmits the emotional state information to the emotion recognition and feedback module. The data communication unit is responsible for data communication between the VR / AR interface module and other modules.

3. According to the artificial intelligence-based traditional culture display system of claim 1, it is characterized in that: The data generation and enhancement module includes a data acquisition unit, a GAN model unit, a data fusion and enhancement unit, and a data evaluation and feedback unit. The data acquisition unit is responsible for acquiring detailed records of user interactions with traditional cultural content and corresponding traditional cultural content information from the big data platform module, and preprocessing the acquired data. The GAN model unit uses the GAN model to generate simulated data similar to real data. The data fusion and enhancement unit fuses the generated simulated data with the real data and enhances the fused data. The data evaluation and feedback unit performs quality evaluation on the generated simulated data and the enhanced data.

4. The traditional culture display system based on artificial intelligence according to claim 1 is characterized in that: The specific steps of the GAN model unit using GAN to generate new virtual data are: Data collection: Collect the existing processed interaction data between users and traditional cultural content; Feature extraction: extract key features related to user interests, preferences, and traditional cultural content characteristics from the preprocessed interaction data; GAN model training: Build a GAN model, including a generator and a discriminator. The task of the generator is to generate virtual data similar to real data, while the task of the discriminator is to distinguish between real data and virtual data. Through training, the generator can generate virtual data that is closer and closer to real data. Simulated data generation: Use the trained generator to generate new simulated data.

5. The traditional culture display system based on artificial intelligence according to claim 4 is characterized in that: In feature extraction, PCA is used to extract key variables that can represent user behavior and traditional cultural content characteristics. The specific steps are: Data standardization: Before performing PCA, the data is standardized by subtracting the mean and dividing by the standard deviation so that the mean of each feature is 0 and the variance is 1. The calculation formula is Among them, X is the original data, μ is the mean, σ is the standard deviation, and Z is the standardized data; Calculate the covariance matrix: The covariance matrix is ​​a matrix that describes the relationship between the various features in the data set. Its elements represent the covariance between different features. The covariance matrix is ​​expressed as Where n is the number of samples, Z T is the transposed matrix of Z; Solve the eigenvalue and eigenvector: Perform eigendecomposition on the covariance matrix C to obtain the eigenvalue λ and the corresponding eigenvector v. The eigenvalue λ represents the variance explained by each principal component, and the eigenvector v represents the direction of each principal component in the original feature space. Select principal components: According to the size of the eigenvalues, select the eigenvectors corresponding to the first k largest eigenvalues, and the eigenvectors constitute a new feature space; Construct the fused feature vector: Project the original feature matrix X onto the new feature space to obtain the fused feature vector.

6. The traditional culture display system based on artificial intelligence according to claim 4 is characterized in that: The specific steps of the GAN model training are: Constructing the GAN model: Generator: Design a neural network structure whose input is a random noise vector and whose output is virtual data similar to the detailed records of the user's interaction with traditional cultural content and the corresponding traditional cultural content information; Discriminator: Another neural network structure whose input is the detailed records of the user's interaction with traditional cultural content and the real data of the corresponding traditional cultural content information or the virtual data generated by the generator, and whose output is the probability that the data is real data; Define the loss function: Generator loss: measures the difference between the virtual data generated by the generator and the real data. The generator loss function is Where z represents a random noise vector, G(z) represents virtual data generated by the generator, and D(G(z)) represents the predicted probability of the discriminator for the virtual data generated by the generator; Discriminator loss: measures the ability of the discriminator to distinguish between real data and virtual data. The discriminator loss function is Among them, x represents the real data, and D(x) represents the predicted probability of the discriminator for the real data; Training process: Initialize model parameters: assign random weights to the generator and discriminator; iterative training: train the discriminator: use real data and virtual data generated by the generator to train the discriminator so that it can accurately distinguish between the two; train the generator: use the feedback from the discriminator to train the generator so that the virtual data it generates is closer and closer to the real data; update weights: after each iteration, update the weights of the generator and discriminator according to the gradient of the loss function; through the iterative training process, continuously optimize the weights of the generator and discriminator so that the generator generates virtual data that is closer and closer to the real data.

7. The traditional culture display system based on artificial intelligence according to claim 1 is characterized in that: The cost-sensitive learning module adopts a cost-sensitive learning method to balance the influence of different categories of samples in the enhanced data in model training, including the following steps: Data preprocessing: Clean, organize and format the user interaction data of the traditional cultural display system, and identify and handle imbalance problems in the data; Weight allocation: Assign weights according to the cost of samples of different categories; Application of cost-sensitive learning algorithm: Introduce weights in the loss function to reflect the cost differences of samples of different categories, and use weighted cross entropy loss function: Among them, w i is the weight of sample i, y i is the true label of sample i, is the predicted label of sample i, and N is the total number of samples; Weight information transmission: transmit weighted data to the machine learning algorithm module.

8. The traditional culture display system based on artificial intelligence according to claim 1 is characterized in that: The machine learning algorithm module includes a data preprocessing unit, a feature selection and dimensionality reduction unit, a model training unit, a model evaluation and optimization unit, and a model deployment and update unit. The data preprocessing unit preprocesses the data obtained from the big data platform module and the data generation and enhancement module. The feature selection and dimensionality reduction unit selects the most valuable features for model training from the preprocessed data and performs dimensionality reduction processing. The model training unit trains the recommendation model based on the selected feature data and the weight information provided by the cost-sensitive learning module. The model evaluation and optimization unit evaluates the trained model and optimizes and adjusts the model according to the evaluation results. The model deployment and update unit deploys the optimized model to the personalized recommendation module for users to make real-time recommendations and interactions.

Images