Method and device for generating multi-modal mixed reality learning resources for embodied agent teaching of primary school mathematics, equipment and storage medium

Abstract

The application discloses a primary school mathematics multi-modal mixed reality learning resource generation method for embodied agent teaching, and realizes semantic linking of a symbol system and an entity environment by mapping a symbolic entity in a question to a physical attribute of an embodied environment through dynamic modeling of a mathematical entity based on a material property character theory; a problem solving step is dynamically converted into multi-modal resources through guided dialogue and situational sketch generation driven by a material property relationship, three-dimensional mathematical entities generated are dynamically embedded in a physical environment through teaching scene modeling and an adaptive mechanism based on Gaussian splashing, and an agent supports a learner in a MR environment. Through a task-scene joint representation model, learning resources and spatial layout are dynamically optimized, and a teaching sequence is adapted to a cognitive state of the learner and environmental characteristics. Thus, a dynamically interactive mixed reality learning environment is provided, and the learner is guided in real time to complete a mathematical task based on self-generated resources, and efficient teaching of human-computer collaboration is realized.

Description

Technical Field

[0001] This application relates to the field of learning resource generation technology, and more specifically, to a method, apparatus, device, and storage medium for generating multimodal mixed reality learning resources for elementary school mathematics for embodied agent teaching. Background Technology

[0002] Embodied learning environments are human-computer collaborative learning environments based on embodied intelligence (such as humanoid robots), aiming to enhance learners' cognitive experience through physical interaction. Symbolic math word problems, as a core resource in elementary school mathematics teaching, emphasize the integration of the subject's symbolic system with cognitive logic. The core challenge of embodied modeling of symbolic math word problems lies in establishing a cross-modal transformation mechanism between the symbolic system and the embodied environment, requiring the dynamic mapping between subject knowledge ontology analysis and cognitive logic. Current technologies rely on predefined templates or limited rule bases, making it difficult to cover diverse question types and adapt to learners' real-time cognitive states. Specifically, the following shortcomings exist:

[0003] First, there is a lack of contextual element mapping: the absence of a dynamic association mechanism between mathematical entities and the embodied physical environment prevents the symbolic system from being effectively transformed into interactive physical entity attributes. Second, there is a cross-modal disconnect: static visual resources are separated from the agent's teaching actions, making it impossible to generate coherent action-semantic chains. Third, there is insufficient context adaptation: the resource generation process does not incorporate the geometric structure of the learning scenario, task objectives, and user cognitive state, resulting in a disconnect between resources and the physical environment, making it difficult to support personalized teaching sequences. Fourth, there is a bottleneck in dynamic interaction: 2D images generated by existing methods are difficult to transform into operable 3D models of mixed reality scenarios, limiting the guiding capabilities of the embodied agent. Summary of the Invention

[0004] To address at least one deficiency or improvement need in the existing technology, this invention provides a method, apparatus, device, and storage medium for generating multimodal mixed reality learning resources for elementary school mathematics in embodied agent teaching. This enables a dynamically interactive mixed reality learning environment, where the embodied agent guides learners to complete mathematical tasks in real time based on self-generated resources, achieving efficient human-computer collaborative teaching.

[0005] To achieve the above objectives, according to a first aspect of the present invention, a method for generating multimodal mixed reality learning resources for elementary school mathematics oriented towards embodied agent teaching is provided, the method comprising:

[0006] S1. Extract mathematical entities from the mathematical problem, and construct the property relationships of the solution process based on the property relationships between the mathematical entities; use the arithmetic operations in the solution steps of the solution process and the property relationships between the mathematical entities to reconstruct the property relationships of quantity state changes; use a dialogue generation model to convert the reconstructed property relationships into natural language question-and-answer pairs to generate guided conversations; and generate a contextual teaching diagram based on the quantity state of the entities in the solution steps.

[0007] S2, capture the environmental features of the physical environment in which the intelligent agent is located to construct the initial learning space of the physical environment, use the contextual teaching diagram and the problem text entity set of mathematical problems to reconstruct the geometric shape of mathematical entities, and enhance the knowledge elements of the geometric shape of the mathematical entities based on the dynamic physical environment, thereby transforming the initial learning space into a mixed reality learning space.

[0008] S3, decompose the learning task into multiple sub-tasks, establish a mapping relationship between the multiple sub-tasks and the mixed reality learning space, establish a task-driven learning context joint representation model based on the mapping relationship; obtain the learning context joint representation decoding parameters corresponding to the current learning scene based on the learning context joint representation model, construct a three-dimensional scene based on the renderer using the learning context joint representation decoding parameters, and project the multimodal resources into the three-dimensional scene to generate multimodal mixed reality learning resources.

[0009] Furthermore, the aforementioned method for generating multimodal mixed reality learning resources for elementary school mathematics aimed at embodied agent teaching also includes:

[0010] S11: Extract mathematical entities from the mathematical problem, establish the property role descriptions of the mathematical entities, and construct the property relationship groups between the mathematical entities based on the property role descriptions to construct the property relationships in the solution process;

[0011] S12, the solution process is decomposed into multiple problem-solving steps, the arithmetic operations in the problem-solving steps are matched and transformed with the material relationships between mathematical entities to reconstruct the material relationships of quantity state changes, and the reconstructed material relationships are transformed into natural language question-and-answer pairs using a dialogue generation model to generate guided conversations.

[0012] S13, Generate a contextual teaching diagram based on the number and status of entities in the problem-solving steps.

[0013] Furthermore, the aforementioned method for generating multimodal mixed reality learning resources for elementary school mathematics aimed at embodied agent teaching also includes:

[0014] The description of the physical properties includes physical attribute description, quantitative relationship description of mathematical entities, and spatial location description;

[0015] Among them, the quantitative relationship description of mathematical entities is used to generate arithmetic expressions for solving problems, while the physical attribute description and spatial location description are used to perform scenario modeling of mathematical problems.

[0016] Furthermore, step S12 in the above-mentioned method for generating multimodal mixed reality learning resources for elementary school mathematics oriented towards embodied agent teaching also includes:

[0017] S121: Input the mathematical problem into the large model to obtain the solution process, and decompose the solution process into multiple problem-solving steps. Each problem-solving step includes the initial state of the quantitative relationship, the final state of the quantitative relationship, and arithmetic operators.

[0018] S122, find the arithmetic operation relation that matches the arithmetic operator in each problem-solving step in the property relation group, and convert the arithmetic operator into a property relation related to the mathematical entity according to the arithmetic operation relation, forming a quantitative state change description based on the property relation:

[0019] S123, using a dialogue generation model, a guiding question and an answer are generated from a description of quantitative state changes based on property relationships to form a question-answer pair.

[0020] Furthermore, step S2 in the above-mentioned method for generating multimodal mixed reality learning resources for elementary school mathematics oriented towards embodied agent teaching also includes:

[0021] S21, the observation sequence is obtained by encoding the multi-view perception sequence of the agent through a pre-trained variational autoencoder, and the predicted potential sequence is obtained based on the latent representation of the defined rendered image. The potential sequence is decoded to obtain the three-dimensional Gaussian ellipsoid parameters, and the initial learning space of the physical environment is constructed by combining the Gaussian ellipsoid parameters of the physical environment objects.

[0022] S22, through cross-modal semantic retrieval, matches the entity most similar to the mathematical entity from the subject resource entity database, uses the most similar entity and its semantic description to generate the corresponding dynamic three-dimensional Gaussian shape, and embeds the dynamic three-dimensional Gaussian shape into the physical environment to construct a mixed reality learning space that integrates the mathematical entity of the question text.

[0023] Furthermore, step S3 in the above-mentioned method for generating multimodal mixed reality learning resources for elementary school mathematics oriented towards embodied agent teaching also includes:

[0024] S31, a joint representation model of the learning context is obtained by performing joint representation modeling on the cognitive elements and the learning context representation of the current scene. The joint representation model of the learning context integrates the user's cognitive preferences and context preferences into a unified semantic space through a multimodal interaction mechanism. The parameters of the joint representation model of the learning context are decoded into Gaussian splash parameters using a preset Gaussian splash optimization objective.

[0025] S32, establish a relational representation between multimodal learning resources and Gaussian splashing scene to represent learning resources, and define the state space by the joint representation of Gaussian splashing scene and learning context;

[0026] S33, update the Gaussian splash parameters using the learning resources and the state space, construct a 3D scene based on the renderer, and project the multimodal resources into the 3D scene to generate multimodal mixed reality learning resources.

[0027] According to a second aspect of the present invention, a multimodal mixed reality learning resource generation device for elementary school mathematics teaching oriented towards embodied intelligent agents is also provided, comprising:

[0028] The learning resource generation module is configured to extract mathematical entities from mathematical problems, construct the property relationships of the solution process based on the property relationships between mathematical entities, reconstruct the property relationships of quantity state changes using arithmetic operations in the solution steps of the solution process and the property relationships between mathematical entities, use a dialogue generation model to convert the reconstructed property relationships into natural language question-and-answer pairs to generate guided conversations, and generate contextual teaching diagrams based on the quantity state of entities in the solution steps.

[0029] The mixed reality learning scenario construction module is configured to capture the environmental features of the physical environment in which the agent is located to construct an initial learning space of the physical environment, reconstruct the geometric shape of mathematical entities using contextual teaching diagrams and a set of mathematical problem text entities, and enhance the knowledge elements of the geometric shape of the mathematical entities based on the dynamic physical environment, thereby transforming the initial learning space into a mixed reality learning space.

[0030] The learning scenario contextualization adaptation module is configured to decompose the learning task into multiple sub-tasks, establish a mapping relationship between the multiple sub-tasks and the mixed reality learning space, establish a task-driven learning scenario joint representation model based on the mapping relationship, obtain the learning scenario joint representation decoding parameters corresponding to the current learning scenario based on the learning scenario joint representation model, construct a three-dimensional scene based on the renderer using the learning scenario joint representation decoding parameters, and project multimodal resources into the three-dimensional scene to generate multimodal mixed reality learning resources.

[0031] According to a third aspect of the present invention, a multimodal mixed reality learning resource generation device for elementary school mathematics oriented towards embodied intelligent agent teaching is also provided, which includes at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit performs the steps of any of the methods described above.

[0032] According to a fourth aspect of the invention, a storage medium is also provided that stores a computer program executable by an access authentication device, which, when run on the access authentication device, causes the access authentication device to perform the steps of any of the methods described above.

[0033] According to a fifth aspect of the present invention, a computer program product is also provided, comprising a computer program / instructions, characterized in that, when executed by a processor, the computer program / instructions implement the steps of any of the methods described above.

[0034] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0035] First, it addresses the cross-modal transformation problem between symbolic and embodied learning: A dynamic modeling mechanism for mathematical entities based on the theory of materiality roles is designed to map symbolic entities in problems to physical attributes of the embodied environment, achieving a semantic link between the symbolic system and the entity environment. Second, it enables the self-generation of multimodal teaching resources: Through guided dialogue driven by materiality relationships and the generation of contextualized diagrams, problem-solving steps are dynamically transformed into multimodal resources, ensuring synchronization between teaching actions and cognitive logic. Third, it constructs adaptive mixed reality scenarios: A teaching scenario modeling and adaptation mechanism based on Gaussian splashing is established, dynamically embedding generated 3D mathematical entities into the physical environment, supporting the agent to guide learners in the MR environment. Fourth, it supports personalized context adaptation: Through a task-scenario joint representation model, learning resources and spatial layout are dynamically optimized, adapting teaching sequences (such as problem-solving steps) to the learner's cognitive state and environmental characteristics. Through the dynamically interactive mixed reality learning environment provided by this invention, embodied agents guide learners to complete mathematical tasks in real time based on self-generated resources, achieving efficient human-computer collaborative teaching. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 A flowchart illustrating a method for generating multimodal mixed reality learning resources for elementary school mathematics in embodied agent teaching, provided for embodiments of this application;

[0038] Figure 2 This is a schematic diagram of the structure of a multimodal mixed reality learning resource generation device for elementary school mathematics oriented towards embodied intelligent agent teaching, provided in an embodiment of this application. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0040] The terms "first," "second," "third," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0041] To achieve the above objectives, according to a first aspect of the present invention, a method for generating multimodal mixed reality learning resources for elementary school mathematics oriented towards embodied agent teaching is provided, the method comprising:

[0042] S1. Extract mathematical entities from the math problem, and construct the property relationships of the solution process based on the property relationships between mathematical entities; reconstruct the property relationships of quantity state changes using arithmetic operations in the solution steps and the property relationships between mathematical entities; use a dialogue generation model to transform the reconstructed property relationships into natural language question-and-answer pairs to generate guided conversations; and generate contextual teaching diagrams based on the quantity states of entities in the solution steps. It should be noted that this multimodal embodied learning resource generation step transforms math word problems into embodied teaching resources, including property relationship modeling, dialogue generation, and diagram generation. The core innovation lies in establishing a dynamic link between symbols and entities through mathematical entity recognition and property role theory, solving the problem of contextual element mapping.

[0043] S2 involves capturing the environmental characteristics of the physical environment in which the agent exists to construct an initial learning space. Using contextualized teaching diagrams and a set of mathematical problem text entities, the geometric shapes of mathematical entities are reconstructed. Based on the dynamic physical environment, knowledge elements are enhanced on these geometric shapes, thereby transforming the initial learning space into a mixed reality learning space. It should be noted that...

[0044] S3, the learning task is decomposed into multiple sub-tasks, and a mapping relationship is established between the multiple sub-tasks and the mixed reality learning space. Based on the mapping relationship, a task-driven learning context joint representation model is established. Based on the learning context joint representation model, the decoding parameters of the learning context joint representation corresponding to the current learning scene are obtained. Using the decoding parameters of the learning context joint representation, a 3D scene is constructed based on the renderer, and multimodal resources are projected into the 3D scene to generate multimodal mixed reality learning resources. It should be noted that this step dynamically optimizes the learning scene according to the environmental and task characteristics, supporting personalized teaching sequences. To achieve this goal, this solution proposes a dynamic generation method for embodied learning scenarios, mainly including contextualized representation of learning scenarios and contextualized adaptation of learning scenarios. Among them, contextualized representation aims to establish a feature joint calculation model between mixed reality learning environment clusters and learning scenarios, providing semantic support for the contextualized adaptation of embodied learning environments. Using the calculated associated features, a dynamic adaptation method for the learning space and a multimodal learning resource adaptation method are established, ultimately forming a contextualized embodied learning environment and resources.

[0045] Furthermore, the aforementioned method for generating multimodal mixed reality learning resources for elementary school mathematics aimed at embodied agent teaching also includes:

[0046] S11 involves extracting mathematical entities from the math problem, establishing property role descriptions for these entities, and constructing property relationship groups between them to build the property relationships in the solution process. It should be noted that mathematical entity property role modeling involves establishing a set of common mathematical entities and associating entities with relevant attributes and example images based on property role theory. The modeling process mainly includes mathematical entity recognition and property role extraction. Specifically, mathematical entity recognition includes detecting and extracting mathematical entities from the problem using a finely tuned named entity recognition model M. ner The set of explicit mathematical entities E in the test question exp At the same time, complete the implicit mathematical entity set E in the problem based on the context. imp denoted as E = {E} exp E imp}←M ner (Text prob ), where Text prob The title text.

[0047] The extraction of physical property relationships of mathematical entities includes: establishing physical property role descriptions of mathematical entities based on physical property role theory, and obtaining groups of physical property relationships between mathematical entities from these descriptions. Specifically, five types of physical property roles are used to describe common physical attributes, quantitative relationships, and spatial locations of mathematical entities. These five types of physical property roles are: Constituent Role (CON), used to describe the part-whole relationship of a mathematical entity or quantity; Unit Role (UNI), used to describe the quantity and unit relationships of a mathematical entity; Material Role (MAT), used to describe the physical material composition of a mathematical entity; Behavioral Role (ACT), used to describe the actions performed on a mathematical entity, with each type of action corresponding to one or more arithmetic operations; and Positioning Role (ORI), used to describe the spatial location relationships of mathematical entities. Among these, the Constituent Role, Unit Role, and Behavioral Role jointly describe the quantitative relationships of mathematical entities, used to generate arithmetic expressions for problem solutions; the Material Role and Positioning Role describe the physical attributes and spatial location information of mathematical entities, suitable for scenario modeling of mathematical problems.

[0048] The modeling of the physical property relationships of mathematical entities relies on a finely tuned annotation model M. qualia The model uses the title text as its basis. prob The set of mathematical entities E extracted in the previous step is used as input, and the set of physical property relationships between the mathematical entities Q is output. ua ={q i =(e i ,e j ,a ij ,r ij )}←M qualia (E). Among them, a ij For entity e i ,e j The implied arithmetic relation, r ij For e i ,e j The material roles between them.

[0049] S12, the solution process is decomposed into multiple problem-solving steps. Arithmetic operations within these steps are matched and transformed with the property relationships between mathematical entities to reconstruct the property relationships of quantity state changes. A dialogue generation model is then used to transform these reconstructed property relationships into natural language question-and-answer pairs to generate guided dialogue. It should be noted that the key to guided teaching dialogue lies in providing problem-solving clues through dialogue. Its core is to transform the problem-solving process into multiple solution clues, enabling learners to dynamically construct the solution process through these clues. Therefore, this step proposes a guided teaching dialogue generation method based on the connection of property relationships.

[0050] S13, Generate a contextualized teaching diagram based on the entity quantity status in the problem-solving steps. It should be noted that, in order to generate the contextualized teaching diagram, we generate it based on the entity quantity status in the problem-solving steps. Generate three scenario diagrams: initial state diagram End state diagram and action diagram And ensure that the generated images meet the constraints of the problem background, mathematical entities, and their quantity changes. To ensure the accuracy and standardization of the legends and action illustrations, we have established a library of entity legends and a library of quantity change action legends for elementary school mathematics, based on the problem-solving steps S. k Entities in and arithmetic operations a ij Schematic symbols are retrieved from the entity symbol library and the action symbol library respectively, and used as input constraints. and The generation process is detailed in Algorithm 1.

[0051]

[0052] In the prompt word generation stage of Algorithm 1, the converter is first used to convert the solution steps S k and quantity state change D k Mapped to descriptive text (Text) respectively sk and Text dk Then, construct three states I in sequence. k ,F k and D k The corresponding contextual image prompts. In the contextual image generation stage, the image generator LLM is first used. image Generate initial state context image To enhance the consistency of the mathematical entities and their quantitative relationships contained in the generated context images, in the generation... and At the same time, the image of the previous state is also used as input. Simultaneously, this scheme further uses a consistency checker (validator) to check the consistency of mathematical entities and their quantitative relationships in the generated image. To improve the ability to understand image content, this scheme uses a dedicated object detector (extrator) to extract mathematical entities and quantitative relationships from the image, and inputs the extraction results into the validator to complete the consistency check. For images that do not meet the consistency requirements, the image generator (LLM) is called again. image .

[0053] Furthermore, the aforementioned method for generating multimodal mixed reality learning resources for elementary school mathematics aimed at embodied agent teaching also includes:

[0054] The description of material properties includes physical attribute descriptions, quantitative relationship descriptions of mathematical entities, and spatial location descriptions. The quantitative relationship descriptions of mathematical entities are used to generate arithmetic expressions for solving problems, while the physical attribute and spatial location descriptions are used for scenario modeling of mathematical problems. It should be noted that, based on the theory of material property roles, a description of the material property roles of mathematical entities is established, and the material property relationship groups between mathematical entities are obtained from these descriptions. Specifically, five types of material property roles are used to describe common physical attributes, quantitative relationships, and spatial locations of mathematical entities. These five types of material property roles are: Constituent Role (CON), used to describe the part-whole relationship of a mathematical entity or quantity; Unit Role (UNI), used to describe the quantity and unit relationship of a mathematical entity; Material Role (MAT), used to describe the physical material composition of a mathematical entity; Behavioral Role (ACT), used to describe the actions performed on the mathematical entity, with each type of action corresponding to one or more arithmetic operations; and Positioning Role (ORI), used to describe the spatial location relationship of a mathematical entity. Among them, the constituent role, unit role, and behavior role jointly describe the quantitative relationships of mathematical entities and are used to generate arithmetic expressions for solving problems; the material role and the positioning role are used to describe the physical attributes and spatial location information of mathematical entities and are suitable for scenario modeling of mathematical problems.

[0055] Furthermore, step S12 in the above-mentioned method for generating multimodal mixed reality learning resources for elementary school mathematics oriented towards embodied agent teaching also includes:

[0056] S121 involves inputting a math problem into a large model to obtain a solution process, which is then broken down into multiple solution steps. Each solution step includes an initial state of the quantitative relationships, a final state of the quantitative relationships, and arithmetic operators. It should be noted that a general large language model is used to generate the solution process, and this process is further broken down into multiple solution steps. Specifically, the problem text (Text) is first... prob Input into large model LLM solver This yields a detailed solution process S. Then, S is broken down into multiple solution steps {S1, S2, ..., S...}. n}, each step S k Includes an initial state I of quantitative relationships k A quantitative relationship ending state F k and an arithmetic operation O k This operation changes the quantitative relationship from the initial state I. k Transition to the final state F k denoted as S k =(I k O k ,F k ), where I k and F kLet O represent the initial and final states at step k, respectively. k Indicates that I k Convert to F k Arithmetic operations.

[0057] S122, in the property relation group, find the arithmetic operation relation that matches the arithmetic operator in each problem-solving step, and convert the arithmetic operator into a property relation related to the mathematical entity according to the arithmetic operation relation, forming a description of the quantity state change based on the property relation: It should be noted that this step mainly realizes the construction of the property relation of quantity state change. In the process of reconstructing the property relation connection, we will solve the problem step S k =(I k O k ,F k Arithmetic operations O in ) k The set of physical property relationships between mathematical entities Q={q i =(e i ,e j ,a ij ,r ij Arithmetic operation relation {a} in} ij Match the steps S and solve the problem. k Converted to mathematical entity e i ,e j Related physical property relationships d ij The implementation process is as follows: First, for each solution step S... k =(I k O k ,F k Find the property relationship with O in the property relationship group Q. k Matching arithmetic operations relation a ij Then, according to a ij S k Converted to mathematical entity e i ,e j Related physical property relationships r ij This forms a description of quantitative state changes based on the relationship between physical properties: in and Representing steps S respectively k Related mathematical entities and property relationships. The construction of property relationships of quantity state changes can match and transform the arithmetic operations in the problem-solving steps with the property relationships between mathematical entities, thereby reconstructing the problem-solving steps to be consistent with the property relationships of mathematical entities, and providing a foundation for the subsequent guided conversation generation.

[0058] S123, using a dialogue generation model, a guiding question and an answer are generated from the description of quantitative state changes based on property relationships to form a question-answer pair. It should be noted that this step primarily achieves guided conversation generation. During guided conversation generation, the dialogue generation model is used to generate a guided question and an answer from the reconstructed description of quantitative state changes D(I). k ←F k This is converted into natural language question-and-answer pairs to guide students step-by-step in solving the problems. Specifically, it involves the number of entities and states in the problem-solving steps. Using the dialogue generation model LLM dialog Generate a guiding question Q k And an answer A k Forming question-and-answer pairs (Q&A) k A k Guiding Question Q k Pointing to key information in the problem-solving steps, such as mathematical entities. and and arithmetic operations Answer A k It includes the process of how to construct a solution using the key information mentioned above.

[0059] Furthermore, step S2 in the above-mentioned method for generating multimodal mixed reality learning resources for elementary school mathematics oriented towards embodied agent teaching also includes:

[0060] S21, the observation sequence is obtained by encoding the agent's multi-view perception sequence through a pre-trained variational autoencoder, and the predicted latent sequence is obtained based on the latent representation of the defined rendered image. The latent sequence is then decoded to obtain the three-dimensional Gaussian ellipsoid parameters, and combined with the Gaussian ellipsoid parameters of the physical environment objects to construct the initial learning space of the physical environment. It should be noted that, in order to capture the geometric structure and appearance features of the physical environment in which the agent is located, this scheme uses a semantic spatial structure generation model based on the agent's multi-view video perception sequence. Specifically, through the agent's multi-view perception sequence... in Represents the view from the k-th viewpoint The output objective is to generate a spatial volume G(b,c,s) with educational semantic labels, where b is the geometric structure of the spatial object, and c∈R. 3 For the appearance color, s∈R d The model employs a cascaded prediction architecture, combining spatiotemporal diffusion and physical constraint optimization. This results in implicitly embedded educational semantic feature vectors.

[0061] First, the agent's multi-view perception sequence V is encoded using a pre-trained variational autoencoder (VAE) to obtain the observation sequence. K is the number of observation frames, and t is the number of denoising steps.

[0062] Define the latent representation of the rendered image as: This is a special case that can be simplified to o. The predicted latent sequence is denoted as .

[0063] Then, a video diffusion model is constructed to generate predicted latent sequences based on the following conditions:

[0064]

[0065] In the formula, θ represents the denoising network parameters, and c represents the text instruction encoding.

[0066] Finally, the diffusion model obtained Decoded into 3D Gaussian ellipsoid parameters g i Including the center coordinates μ i Covariance matrix Σ i spherical harmonic coefficient SH i Transparency α i The 3D Gaussian ellipsoidal parameters of all physical environment objects constitute the initial spatial shape G. p From the initial spatial shape G p The resulting 3D space is denoted as Γ0.

[0067] S22 involves using cross-modal semantic retrieval to match the entity most similar to the mathematical entity from the subject resource entity database. The most similar entity and its semantic description are then used to generate a dynamic 3D Gaussian shape corresponding to the entity. This dynamic 3D Gaussian shape is embedded into the physical environment to construct a mixed reality learning space that integrates the mathematical entities from the problem text. It should be noted that semantically associating various elements in the scene with embodied learning resources enables learners to acquire richer and more targeted learning content within specific learning contexts. Based on the initial learning space Γ0, the subject resource entity database K, and the problem text entity set E, a mixed reality learning scene G with strong educational semantics is formed. e This ensures a strong connection between spatial elements and educational semantics. To this end, this solution employs a two-step strategy: reconstruction of mathematical entity geometric shapes and enhancement of geometric shape knowledge elements, to achieve the transformation from Γ0 to the mixed reality learning space Γ. e The leap.

[0068] First, mathematical entity geometric shape reconstruction: targeting the entity set of the problem text. mathematical entity e in i The system uses cross-modal semantic retrieval to match the most similar entity k from the subject resource entity database K. e ∈K (i.e.) in (For text encoders). (k) e and its semantic description Input a conditional 3D Gaussian generative model, output a dynamic 3D Gaussian shape corresponding to the entity. Each Gaussian element g j Parameterized as (μ) j ,Σ j ,SH j ,α j ). Through a semantically guided registration module, G t Adapted to the physical environment model G p middle:

[0069] Position alignment: Calculate the associated position of entities (MLP outputs offset, GAP is global average pooling);

[0070] Form optimization: Adjust Σ j Scaling factor makes G t With G p Consistent spatial scale;

[0071] Semantic binding: incorporating educational semantic tags Inject SH j The color channels enable the integration of geometry and semantics.

[0072] The above adaptation algorithm will be used to adapt G t Embedded into G p This creates a mixed reality learning space that integrates problem texts and mathematical entities. e .

[0073] Then, the knowledge elements of geometric shapes are enhanced. Based on the physical environment model. Gaussian points were divided into clusters C using a spatial density clustering algorithm. i Extracting cluster geometric features f geo (C i The optimal entity is retrieved from the subject resource entity database K. satisfy Will Geometric semantics (s) geo ), physical semantics (s phy ), knowledge semantics (s knol Injection into cluster C i Educational semantic enhancement is achieved by incorporating Gaussian parameters into the geometric shape. The specific execution process is shown in Algorithm 2:

[0074]

[0075] Algorithm 2 first processes the input physical environment model G p Geometric shape partitioning was performed, and the Gaussian point cloud was divided into clusters C using the DBSCAN clustering algorithm. i And extract the cluster geometric features fgeo (C i The key is used as the retrieval key. Then, the optimal entity is matched from the subject resource entity database K through cross-modal semantic retrieval. Ensure that geometric features are aligned with educational semantics. Then, assemble the entity. The multidimensional semantics (including geometric, physical, and knowledge attributes) are encoded into feature vectors and injected with Gaussian point parameters. Finally, in the environment reconstruction stage, a new point cloud model is generated based on the semantically enhanced parameters. And output semantically enhanced mixed reality learning scenes through the rendering engine. e The algorithm ensures the accuracy of semantic injection and environmental consistency through dynamic parameterization control. Outliers that are not clustered retain their original parameters to maintain environmental integrity, providing structured support for downstream teaching and interactive tasks.

[0076] Furthermore, step S3 in the above-mentioned method for generating multimodal mixed reality learning resources for elementary school mathematics oriented towards embodied agent teaching also includes:

[0077] S31, a joint representation model of the learning context is obtained by jointly representing the cognitive elements and the current scene. This model integrates user cognitive preferences and contextual preferences into a unified semantic space through a multimodal interaction mechanism. The parameters of the joint representation model are decoded into Gaussian splash parameters using a preset Gaussian splash optimization objective. It should be noted that task decomposition and goal mapping break down complex learning tasks into multiple sub-tasks and establish a mapping relationship between each sub-task and the primitives of the mixed reality learning environment, providing clear direction and goal orientation for the design of learning activities. To this end, we introduce a task-goal mapping function to map the set of sub-tasks... The operation steps, knowledge points, and other attributes in the t i Related to the Gaussian element g in the model j The mapping process is defined as follows:

[0078]

[0079] in, G is the set of decomposed subtasks. p This is a physical environment model. Based on the task-driven, goal-oriented dynamic adaptation principle in embodied cognition theory, the learning task sequence is... With Gaussian metasequence By incorporating a joint feature computation model, dynamic encoding is generated.

[0080] g t =Encoder(T,G p )

[0081] Where g embIt encodes task-driven, goal-oriented features to provide semantic support for context adaptation.

[0082] Based on the above representation model, a system is constructed that includes five categories of educational semantic primitives, namely, the spatiotemporal context primitive c. p Educational entity basic element e p Behavioral pattern primitive b p Target level primitive g p relation attribute primitive r p Joint computation function Based on this, and combining the encoding results of the task objective layer, a task-driven joint representation model of the learning context is established:

[0083]

[0084] Where W is the weight parameter and ⊙ is the feature combination operator.

[0085] S32 establishes a relational representation between multimodal learning resources and the Gaussian splashing scene to characterize the learning resources. The state space is jointly defined by the Gaussian splashing scene and the learning context. It should be noted that, in order to improve the context relevance of the learning space, the elements in the learning space need to be reorganized and adapted. A visual image-based 3D scene modeling framework has been established in research related to the reconstruction of the embodied agent's operational space. Based on this framework, this scheme carries out cognitive-driven dynamic optimization of the 3D scene, as follows:

[0086] First, a joint representation model is performed on the cognitive elements and the learning context representation of the current scene. The model integrates user cognitive preferences (C) and contextual preferences (V) into a unified semantic space R through a multimodal interaction mechanism. cj =MLP(C⊕V).

[0087] Then, set the Gaussian splash optimization target:

[0088]

[0089] In the formula D JS For Jensen-Shannon divergence, force the scene distribution Q(G) s Alignment Educational Semantic Priors P(R) cv Alignment, Cover(G) s G p ) is used to measure the splash scene G s For the physical elementary G p The coverage rate is denoted by λ, where λ is a balancing factor. Based on the optimization objective, the learning context will be jointly represented. Decoded into Gaussian splash parameters:

[0090]

[0091] S33, update the Gaussian splash parameters using the learning resources and the state space, construct a 3D scene based on the renderer, and project the multimodal resources into the 3D scene to generate multimodal mixed reality learning resources. It should be noted that a correlation calculation mechanism is established between contextual features and multimodal resources to achieve accurate dynamic adaptation of multimodal learning resources.

[0092] In terms of learning resource representation, a multimodal learning resource C is established. m Association Representation Mechanism with Gaussian Splash Scene

[0093] In the state-space representation, the Gaussian splash scene G... s Joint representation with learning context Commonly defined state f geo (·) represents the scene feature extraction network, and sets the splash consistency reward function.

[0094] Construct a joint optimization objective:

[0095]

[0096] In the formula, β is the reward factor. Based on joint loss Update Gaussian splash parameters Finally, the 3D scene is output using the Gaussian splash renderer. and multimodal resources Projected onto the corresponding splash location in the scene to achieve the overlay of learning resources.

[0097] As can be seen, this invention dynamically models mathematical entities based on the theory of materiality roles, mapping symbolic entities in problems to physical attributes of the embodied environment, thus achieving a semantic link between the symbolic system and the physical environment. Through guided dialogue driven by materiality relationships and the generation of contextualized diagrams, problem-solving steps are dynamically transformed into multimodal resources. Furthermore, through a teaching scenario modeling and adaptation mechanism based on Gaussian splashing, the generated 3D mathematical entities are dynamically embedded into the physical environment, supporting the agent in guiding learners within the MR environment. A task-scenario joint representation model dynamically optimizes learning resources and spatial layout, ensuring the teaching sequence adapts to the learner's cognitive state and environmental characteristics. This invention, through the organic combination of these three steps, outputs a dynamic and interactive learning environment, enabling the embodied agent to use generated dialogues and resources to guide learners in completing mathematical tasks within a mixed reality scenario, adaptively optimizing the teaching sequence.

[0098] According to a second aspect of the present invention, a multimodal mixed reality learning resource generation device for elementary school mathematics teaching oriented towards embodied intelligent agents is also provided, comprising:

[0099] The learning resource generation module is configured to extract mathematical entities from mathematical problems, construct the property relationships of the solution process based on the property relationships between mathematical entities, reconstruct the property relationships of quantity state changes using arithmetic operations in the solution steps of the solution process and the property relationships between mathematical entities, use a dialogue generation model to convert the reconstructed property relationships into natural language question-and-answer pairs to generate guided conversations, and generate contextual teaching diagrams based on the quantity state of entities in the solution steps.

[0100] The mixed reality learning scenario construction module is configured to capture the environmental features of the physical environment in which the agent is located to construct an initial learning space of the physical environment, reconstruct the geometric shape of mathematical entities using contextual teaching diagrams and a set of mathematical problem text entities, and enhance the knowledge elements of the geometric shape of the mathematical entities based on the dynamic physical environment, thereby transforming the initial learning space into a mixed reality learning space.

[0101] The learning scenario contextualization adaptation module is configured to decompose the learning task into multiple sub-tasks, establish a mapping relationship between the multiple sub-tasks and the mixed reality learning space, establish a task-driven learning scenario joint representation model based on the mapping relationship, obtain the learning scenario joint representation decoding parameters corresponding to the current learning scenario based on the learning scenario joint representation model, construct a three-dimensional scene based on the renderer using the learning scenario joint representation decoding parameters, and project multimodal resources into the three-dimensional scene to generate multimodal mixed reality learning resources.

[0102] The steps performed by each module in the above-described elementary school mathematics multimodal mixed reality learning resource generation device for embodied agent teaching correspond to the steps of the aforementioned elementary school mathematics multimodal mixed reality learning resource generation method for embodied agent teaching, and will not be elaborated here.

[0103] According to a third aspect of the present invention, a multimodal mixed reality learning resource generation device for elementary school mathematics oriented towards embodied intelligent agent teaching is also provided, which includes at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit performs the steps of any of the methods described above.

[0104] According to a fourth aspect of the invention, a storage medium is also provided that stores a computer program executable by an access authentication device, which, when run on the access authentication device, causes the access authentication device to perform the steps of any of the methods described above.

[0105] According to a fifth aspect of the present invention, a computer program product is also provided, comprising a computer program / instructions, characterized in that, when executed by a processor, the computer program / instructions implement the steps of any of the methods described above.

[0106] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0107] First, it addresses the cross-modal transformation problem between symbolic and embodied learning: A dynamic modeling mechanism for mathematical entities based on the theory of materiality roles is designed to map symbolic entities in problems to physical attributes of the embodied environment, achieving a semantic link between the symbolic system and the entity environment. Second, it enables the self-generation of multimodal teaching resources: Through guided dialogue driven by materiality relationships and the generation of contextualized diagrams, problem-solving steps are dynamically transformed into multimodal resources, ensuring synchronization between teaching actions and cognitive logic. Third, it constructs adaptive mixed reality scenarios: A teaching scenario modeling and adaptation mechanism based on Gaussian splashing is established, dynamically embedding generated 3D mathematical entities into the physical environment, supporting the agent to guide learners in the MR environment. Fourth, it supports personalized context adaptation: Through a task-scenario joint representation model, learning resources and spatial layout are dynamically optimized, adapting teaching sequences (such as problem-solving steps) to the learner's cognitive state and environmental characteristics. Through the dynamically interactive mixed reality learning environment provided by this invention, embodied agents guide learners to complete mathematical tasks in real time based on self-generated resources, achieving efficient human-computer collaborative teaching.

[0108] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method. The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, as well as magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.

[0109] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0110] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0111] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0112] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0113] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0114] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0115] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0116] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of embodiments of this disclosure upon considering the specification and practicing the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

[0117] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0118] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating multimodal mixed reality learning resources for elementary school mathematics teaching oriented towards embodied intelligent agents, characterized in that, include: S1, extract mathematical entities from the math problem, and construct the property relationships of the solution process based on the property relationships between the mathematical entities; The arithmetic operations and the property relationships between mathematical entities in the problem-solving steps of the solution process are used to reconstruct the property relationships of quantity state changes. The reconstructed property relationships are transformed into natural language question-and-answer pairs using a dialogue generation model to generate guided conversations. Contextual teaching diagrams are generated based on the entity quantity states in the problem-solving steps. S2, capture the environmental features of the physical environment in which the intelligent agent is located to construct the initial learning space of the physical environment, use the contextual teaching diagram and the problem text entity set of mathematical problems to reconstruct the geometric shape of mathematical entities, and enhance the knowledge elements of the geometric shape of the mathematical entities based on the dynamic physical environment, thereby transforming the initial learning space into a mixed reality learning space. S3, decompose the learning task into multiple sub-tasks, establish a mapping relationship between the multiple sub-tasks and the mixed reality learning space, and establish a task-driven learning context joint representation model based on the mapping relationship; obtain the learning context joint representation decoding parameters corresponding to the current learning scene based on the learning context joint representation model, use the learning context joint representation decoding parameters to construct a three-dimensional scene based on the renderer, and project the multimodal resources into the three-dimensional scene to generate multimodal mixed reality learning resources; Step S3 further includes: S31, a joint representation model of the learning context is obtained by performing joint representation modeling on the cognitive elements and the learning context representation of the current scene. The joint representation model of the learning context integrates the user's cognitive preferences and context preferences into a unified semantic space through a multimodal interaction mechanism. The parameters of the joint representation model of the learning context are decoded into Gaussian splash parameters using a preset Gaussian splash optimization objective. S32, establish a relational representation between multimodal learning resources and Gaussian splashing scene to represent learning resources, and define the state space by the joint representation of Gaussian splashing scene and learning context; S33, update the Gaussian splash parameters using the learning resources and the state space, construct a 3D scene based on the renderer, and project the multimodal resources into the 3D scene to generate multimodal mixed reality learning resources.

2. The method for generating multimodal mixed reality learning resources for primary school mathematics as described in claim 1, characterized in that, Also includes: S11: Extract mathematical entities from the mathematical problem, establish the property role descriptions of the mathematical entities, and construct the property relationship groups between the mathematical entities based on the property role descriptions to construct the property relationships in the solution process; S12, the solution process is decomposed into multiple problem-solving steps, the arithmetic operations in the problem-solving steps are matched and transformed with the material relationships between mathematical entities to reconstruct the material relationships of quantity state changes, and the reconstructed material relationships are transformed into natural language question-and-answer pairs using a dialogue generation model to generate guided conversations. S13, Generate a contextual teaching diagram based on the number and status of entities in the problem-solving steps.

3. The method for generating multimodal mixed reality learning resources for primary school mathematics as described in claim 2, characterized in that, Step S11 further includes: The description of the physical properties includes physical attribute description, quantitative relationship description of mathematical entities, and spatial location description; Among them, the quantitative relationship description of mathematical entities is used to generate arithmetic expressions for solving problems, while the physical attribute description and spatial location description are used to perform scenario modeling of mathematical problems.

4. The method for generating multimodal mixed reality learning resources for primary school mathematics as described in claim 2, characterized in that, Step S12 further includes: S121: Input the mathematical problem into the large model to obtain the solution process, and decompose the solution process into multiple problem-solving steps. Each problem-solving step includes the initial state of the quantitative relationship, the final state of the quantitative relationship, and arithmetic operators. S122, find the arithmetic operation relation that matches the arithmetic operator in each problem-solving step in the property relation group, and convert the arithmetic operator into a property relation related to the mathematical entity according to the arithmetic operation relation, forming a quantitative state change description based on the property relation: S123, using a dialogue generation model, a guiding question and an answer are generated from a description of quantitative state changes based on property relationships to form a question-answer pair.

5. The method for generating multimodal mixed reality learning resources for primary school mathematics as described in claim 1, characterized in that, Step S2 further includes: S21, the observation sequence is obtained by encoding the multi-view perception sequence of the agent through a pre-trained variational autoencoder, and the predicted potential sequence is obtained based on the latent representation of the defined rendered image. The potential sequence is decoded to obtain the three-dimensional Gaussian ellipsoid parameters, and the initial learning space of the physical environment is constructed by combining the Gaussian ellipsoid parameters of the physical environment objects. S22, through cross-modal semantic retrieval, matches the entity most similar to the mathematical entity from the subject resource entity database, uses the most similar entity and its semantic description to generate the corresponding dynamic three-dimensional Gaussian shape, and embeds the dynamic three-dimensional Gaussian shape into the physical environment to construct a mixed reality learning space that integrates the mathematical entity of the question text.

6. A multimodal mixed reality learning resource generation device for elementary school mathematics teaching oriented towards embodied intelligent agents, characterized in that, include: The learning resource generation module is configured to extract mathematical entities from mathematical problems and construct the property relationships of the solution process based on the property relationships between mathematical entities. The arithmetic operations and the property relationships between mathematical entities in the problem-solving steps of the solution process are used to reconstruct the property relationships of quantity state changes. The reconstructed property relationships are transformed into natural language question-and-answer pairs using a dialogue generation model to generate guided conversations. Contextual teaching diagrams are generated based on the entity quantity states in the problem-solving steps. The mixed reality learning scenario construction module is configured to capture the environmental features of the physical environment in which the agent is located to construct an initial learning space of the physical environment, reconstruct the geometric shape of mathematical entities using contextual teaching diagrams and a set of mathematical problem text entities, and enhance the knowledge elements of the geometric shape of the mathematical entities based on the dynamic physical environment, thereby transforming the initial learning space into a mixed reality learning space. The learning scenario contextualization adaptation module is configured to decompose the learning task into multiple sub-tasks, establish a mapping relationship between the multiple sub-tasks and the mixed reality learning space, establish a task-driven learning scenario joint representation model based on the mapping relationship, obtain the learning scenario joint representation decoding parameters corresponding to the current learning scenario based on the learning scenario joint representation model, construct a three-dimensional scene based on the renderer using the learning scenario joint representation decoding parameters, and project multimodal resources into the three-dimensional scene to generate multimodal mixed reality learning resources. The learning scenario contextual adaptation module further includes: A learning context joint representation model is obtained by performing joint representation modeling on the cognitive elements and the learning context representation of the current scene. The learning context joint representation model integrates user cognitive preferences and context preferences into a unified semantic space through a multimodal interaction mechanism. The parameters of the learning context joint representation model are decoded into Gaussian splash parameters using a preset Gaussian splash optimization objective. Establish a relational representation between multimodal learning resources and Gaussian splashing scene to represent the learning resources, and define the state space by the joint representation of Gaussian splashing scene and learning context; The Gaussian splash parameters are updated using the learning resources and the state space, and a 3D scene is constructed based on the renderer. The multimodal resources are then projected into the 3D scene to generate multimodal mixed reality learning resources.

7. A device for generating multimodal mixed reality learning resources for elementary school mathematics teaching to embodied intelligent agents, characterized in that, It includes at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program that, when executed by the processing unit, causes the processing unit to perform the steps of the method according to any one of claims 1 to 5.

8. A storage medium, characterized in that, It stores a computer program executable by an access authentication device, which, when run on the access authentication device, causes the access authentication device to perform the steps of the method according to any one of claims 1 to 5.

9. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 5.

Images