Multifunctional video conference interaction method, device, equipment and storage medium

Abstract

The present application relates to the technical field of data interaction, and discloses a multifunctional video conference interaction method, device, equipment and storage medium. The method includes: collecting the conference theme, setting the conference parameters according to the conference theme, and authenticating the conference parameters to obtain an identity identifier set; collecting historical behavior data, and inputting the conference parameters and historical behavior data into a preset graph neural network to analyze the conference data to obtain a data distribution list; collecting audio and video data in real time, and performing adaptive transmission data analysis on the audio and video data according to the data distribution list to obtain a transmission parameter set; inputting the transmission data set into a preset Seq2Seq model to extract conference data to obtain structured conference record data; constructing a management strategy for the structured conference record data to obtain a conference management strategy, and controlling the video conference in real time through the conference management strategy. The present application improves the real-time and accuracy of multifunctional video conference interaction.

Description

Technical Field

[0001] The present application relates to the field of data interaction, and in particular to a multifunctional video conferencing interaction method, device, equipment and storage medium. Background Art

[0002] Currently, video conferencing technology has been widely used in enterprises and organizations for remote communication and collaboration. Existing video conferencing systems usually have basic video and audio transmission functions, and support interactive functions such as screen sharing, file sharing, and real-time chat. These systems can meet basic meeting needs, provide convenient communication channels, and greatly improve work efficiency and collaboration effects.

[0003] However, existing video conferencing systems still have many deficiencies in interactivity and intelligence. For example, most systems cannot provide personalized material recommendations and management strategies based on the specific content of the meeting and the behavior of the participants. In addition, existing systems lack intelligent means for the generation and management of meeting records, and usually rely on manual recording, which is inefficient and prone to errors. These deficiencies have seriously affected the user experience and management efficiency of video conferencing. Summary of the invention

[0004] The present application provides a multifunctional video conference interaction method, apparatus, device and storage medium for improving the real-time performance and accuracy of multifunctional video conference interaction.

[0005] In a first aspect, the present application provides a multifunctional video conference interaction method, the multifunctional video conference interaction method comprising: collecting a conference theme, and setting conference parameters according to the conference theme, wherein the conference parameters include: basic parameters and participant identity information, and authenticating the conference parameters to obtain an identity identifier set;

[0006] Based on the participant identity information, historical behavior data is collected, and the conference parameters and the historical behavior data are input into a preset graph neural network to perform conference data analysis to obtain a data distribution list;

[0007] Collect audio and video data in real time, and perform adaptive transmission data analysis on the audio and video data according to the data distribution list to obtain a transmission parameter set;

[0008] Inputting the transmission data set into a preset Seq2Seq model to extract conference data to obtain structured conference record data;

[0009] A management strategy is constructed for the structured conference record data to obtain a conference management strategy, and the video conference is controlled in real time through the conference management strategy.

[0010] In combination with the first aspect, in a first implementation of the first aspect of the present application, the conference topic is collected, and conference parameters are set according to the conference topic, wherein the conference parameters include: basic parameters and participant identity information, and the conference parameters are authenticated to obtain an identity identifier set, including:

[0011] Collecting the conference theme and setting the conference parameters according to the conference theme, wherein the conference parameters include: basic parameters and participant identity information;

[0012] Extracting the meeting time, meeting location and meeting agenda from the basic parameters;

[0013] Extracting the participant's identity information to obtain basic information of the participant, wherein the basic information of the participant includes: name, email address and position;

[0014] Constructing a virtual identity image based on the basic information of the participant to obtain multiple virtual identity images;

[0015] Authentication is performed on each of the virtual identity images. When each of the virtual identity images passes the authentication, an identifier is constructed for each of the virtual identity images to obtain the identity identifier set.

[0016] In combination with the first aspect, in a second implementation of the first aspect of the present application, the historical behavior data is collected based on the participant identity information, and the conference parameters and the historical behavior data are input into a preset graph neural network for conference data analysis to obtain a data distribution list, including:

[0017] Collecting historical behavior data based on the participant identity information, wherein the historical behavior data includes: meeting participation records, speech data, and document access data;

[0018] Constructing behavior labels for the participant identity information based on the historical behavior data to obtain multiple groups of behavior labels;

[0019] Constructing behavior nodes for each group of behavior tags respectively to obtain a behavior node set corresponding to each group of behavior tags;

[0020] Inputting the conference parameters and the behavior node set into the graph neural network to construct a conference behavior graph structure, thereby obtaining a target conference behavior graph structure;

[0021] Performing behavior association data analysis on the target meeting behavior graph structure to obtain a behavior association data set;

[0022] The meeting data is matched with the behavior association data set through the basic information of the participants to obtain the data distribution list.

[0023] In combination with the first aspect, in a third implementation of the first aspect of the present application, the real-time collection of audio and video data, and the adaptive transmission data analysis of the audio and video data according to the data distribution list to obtain a transmission parameter set include:

[0024] Collect audio and video data in real time, perform parameter analysis on the audio and video data, and obtain resolution, frame rate, bit rate, and audio sampling rate;

[0025] Performing semantic segmentation on the audio and video data to obtain semantic labels corresponding to multiple groups of sub-audio and video data;

[0026] Segmenting the audio and video data using the data distribution list and the semantic tags corresponding to the multiple groups of sub-audio and video data to obtain multiple groups of target sub-audio and video data;

[0027] Based on the resolution, frame rate, bit rate and audio sampling rate, transmission data analysis is performed on each group of target sub-audio and video data to obtain transmission parameter data corresponding to each group of target sub-audio and video data, and the transmission parameter data corresponding to each group of target sub-audio and video data are merged into the transmission parameter set.

[0028] In combination with the first aspect, in a fourth implementation of the first aspect of the present application, the step of inputting the transmission data set into a preset Seq2Seq model to extract conference data to obtain structured conference record data includes:

[0029] Extracting video data from the transmission data set to obtain key frame data and metadata, wherein the metadata includes: a timestamp and speaker identification data;

[0030] Extracting audio data from the transmission data set to obtain an audio data set, and converting the audio data set into initial text data;

[0031] Performing data cleaning on the initial text data to obtain target text data;

[0032] Performing sequence conversion on the target text data to obtain a word vector set;

[0033] Performing additional feature analysis on the word vector set, the key frame data, and the metadata to obtain an additional feature set corresponding to the word vector set;

[0034] Performing feature fusion on the word vector set and the additional feature set to obtain a target word vector set;

[0035] The target word vector set is input into a preset Seq2Seq model to extract conference data to obtain structured conference record data.

[0036] In combination with the first aspect, in a fifth implementation of the first aspect of the present application, the Seq2Seq model includes an encoder and a decoder, and the target word vector set is input into a preset Seq2Seq model to extract conference data to obtain structured conference record data, including:

[0037] Inputting the target word vector set into the encoder for sequence analysis to obtain context vector representation data;

[0038] Performing sequence conversion on the context vector representation data to obtain context sequence data;

[0039] Inputting the context sequence data into the decoder for structured parameter analysis to obtain a structured parameter set, wherein the structured parameter set includes: paragraphs, punctuation marks, and titles;

[0040] The context sequence data is converted into structured data using the structured parameter set to obtain the structured meeting record data.

[0041] In combination with the first aspect, in a sixth implementation of the first aspect of the present application, the construction of a management strategy for the structured conference record data to obtain a conference management strategy, and the real-time control of the video conference through the conference management strategy includes:

[0042] Performing text segmentation on the structured conference record data to obtain a segmentation set;

[0043] Performing behavior pattern recognition on the word segmentation set to obtain behavior pattern data, wherein the behavior pattern data includes: speech frequency, speech duration, and interaction duration;

[0044] Constructing conference rules through the behavior pattern data to obtain conference rule data;

[0045] Generate a conference management strategy through the conference rule data, and generate a conference management prompt word according to the conference management strategy;

[0046] The video conference is controlled in real time based on the conference management strategy and the conference management prompt words.

[0047] In a second aspect, the present application provides a multifunctional video conference interaction device, the multifunctional video conference interaction device comprising:

[0048] A collection module, used to collect conference topics and set conference parameters according to the conference topics, wherein the conference parameters include: basic parameters and participant identity information, and perform identity authentication on the conference parameters to obtain an identity identifier set;

[0049] An input module, used to collect historical behavior data based on the participant identity information, and input the conference parameters and the historical behavior data into a preset graph neural network to perform conference data analysis to obtain a data distribution list;

[0050] An analysis module, used for collecting audio and video data in real time, and performing adaptive transmission data analysis on the audio and video data according to the data distribution list to obtain a transmission parameter set;

[0051] An extraction module, used for inputting the transmission data set into a preset Seq2Seq model to extract conference data and obtain structured conference record data;

[0052] The construction module is used to construct a management strategy for the structured conference record data, obtain a conference management strategy, and control the video conference in real time through the conference management strategy.

[0053] The third aspect of the present application provides a multifunctional video conferencing interaction device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the multifunctional video conferencing interaction device executes the above-mentioned multifunctional video conferencing interaction method.

[0054] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned multifunctional video conferencing interaction method.

[0055] In the technical solution provided by the present application, the security of the meeting and the authenticity of the participant's identity are greatly improved by using generative adversarial networks (GANs) for participant identity authentication. Generative adversarial networks can generate highly realistic virtual identity images, which can effectively prevent identity theft and forgery by comparing them with the identity information provided by the participants, thereby ensuring the legitimacy of each participant. This high-level identity authentication not only protects the confidentiality of the meeting, but also enhances the trust of the participants. Secondly, the present invention uses graph neural networks (GNNs) to conduct in-depth analysis of meeting parameters and participant historical behavior data to generate a personalized meeting material distribution list. This process ensures that each participant obtains the most relevant meeting materials before the meeting by analyzing the meeting topic, agenda and participant historical behavior data (such as speech records and material access records of previous meetings), thereby improving the efficiency and quality of meeting preparation. This personalized material distribution not only enhances the meeting experience of the participants, but also ensures the pertinence and effectiveness of the meeting discussion. In addition, during the real-time audio and video transmission process, the present invention adopts an adaptive transmission control system based on reinforcement learning. The system can dynamically adjust transmission parameters (such as resolution, frame rate, bit rate, etc.) to adapt to network conditions and user needs, thereby ensuring the stability and quality of transmission. Through the reinforcement learning algorithm, the system can continuously learn and optimize the transmission strategy, and even in the case of network bandwidth fluctuations, it can provide high-quality audio and video transmission services to ensure the smooth progress of the meeting. Furthermore, the present invention introduces a Seq2Seq model based on an attention mechanism for intelligent generation of meeting records. The model can automatically convert the voice content in the meeting into structured meeting records, greatly reducing the workload and error rate of manual recording. The Seq2Seq model processes long segments of speech through an encoder-decoder structure and focuses on key content through an attention mechanism to generate accurate and coherent meeting records. This function not only improves the accuracy and completeness of the meeting records, but also facilitates the review and follow-up of the meeting content after the meeting. The present invention constructs an intelligent meeting management strategy through behavioral pattern recognition and text analysis. The system can monitor the progress of the meeting in real time, analyze the frequency of speech, speech duration and interaction, and generate real-time feedback reports and optimized management strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0057] Figure 1 A schematic diagram of an embodiment of a multifunctional video conferencing interaction method in an embodiment of the present application;

[0058] Figure 2 It is a schematic diagram of an embodiment of a multifunctional video conferencing interactive device in an embodiment of the present application. DETAILED DESCRIPTION

[0059] Embodiments of the present application provide a multifunctional video conference interaction method, device, equipment and storage medium. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0060] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the multifunctional video conference interaction method in the embodiment of the present application includes:

[0061] Step S101: Collect the conference theme and set the conference parameters according to the conference theme, wherein the conference parameters include: basic parameters and participant identity information, and perform identity authentication on the conference parameters to obtain an identity identifier set;

[0062] It is understandable that the execution subject of the present application may be a multifunctional video conference interactive device, or a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0063] Specifically, the meeting theme is collected through a pre-set channel (such as email, meeting notification system, etc.), and meeting parameters are set according to the meeting theme, including basic information of the meeting, such as meeting time, meeting place and meeting agenda, as well as identity information of participants, such as name, email address and position. When extracting basic parameters, the time, place and agenda of the meeting are automatically identified and extracted, and this information is stored in the meeting parameter database. When extracting the identity information of the participants, the basic information of each participant is extracted to obtain basic information including name, email address and position. Based on the basic information, a virtual identity image is constructed for each participant. The construction process of the virtual identity image includes generating a digital virtual image using the participant's photo or other identification data. These images can be two-dimensional avatars or three-dimensional full-body images, depending on the implementation method of the system and the needs of the meeting. Each virtual identity image is authenticated to ensure that the identity of the participant is consistent with the information provided. The identity authentication process can be carried out in a variety of ways, including facial recognition technology, fingerprint recognition or other biometric recognition technology. When each virtual identity image passes the identity authentication, a unique identifier is generated for each virtual identity image. These identifiers are not only used for identity identification during the meeting, but also for tracking and management of subsequent meeting records, material distribution and other links. The resulting set of identifiers is stored in a secure database and used in real time during the meeting to ensure that each participant's identity is correctly identified and managed.

[0064] Step S102: Collect historical behavior data based on the participant identity information, and input the conference parameters and historical behavior data into a preset graph neural network to analyze the conference data to obtain a data distribution list;

[0065] Specifically, historical behavior data is collected based on the identity information of the participants, including meeting participation records, speech data, and document access data. Through these data, the meeting behavior habits and preferences of each participant are fully understood. Behavior labels are constructed for the identity information of the participants based on the historical behavior data. Behavior labels can be multi-dimensional, including participation frequency, number of speeches, depth of document reading, etc. These labels can describe the behavioral characteristics of each participant in more detail. Behavior nodes are constructed for each group of behavior labels. Behavior nodes are the specific manifestation of behavior labels. Each node represents the specific behavior performance of a participant in a certain aspect, such as the number of speeches of a participant in a certain meeting or the frequency of accessing documents in a certain period of time. By constructing nodes for each group of behavior labels, a set of behavior nodes corresponding to each group of behavior labels is obtained. These node sets are the basic data for graph neural network analysis. The meeting parameters and the set of behavior nodes are input into the preset graph neural network model to construct the structure of the meeting behavior graph. The graph neural network uses the relationship and connection method between nodes to construct the target meeting behavior graph structure. The graph structure can show the complex relationship between participants and the content of the meeting. The behavior association data analysis is performed on the target meeting behavior graph structure to identify and mine the association between behavior nodes to form a behavior association data set. The associated data can reveal the behavioral trends of participants in the meeting, such as which participants are more inclined to interact, and which participants pay more attention to reading and using materials. The behavioral association data set is matched with the basic information of the participants to obtain the final material distribution list. The matching process needs to combine the behavioral characteristics of the participants and the theme of the meeting to ensure that the distributed materials are of practical value and pertinence to the participants. The material distribution list can not only improve the utilization efficiency of meeting materials, but also make personalized recommendations based on the needs and interests of the participants, thereby improving the interactive effect of the meeting and the satisfaction of the participants.

[0066] Step S103, collecting audio and video data in real time, and performing adaptive transmission data analysis on the audio and video data according to the data distribution list to obtain a transmission parameter set;

[0067] Specifically, the audio and video data generated during the meeting are collected in real time, including key parameters such as video resolution, frame rate, bit rate, and audio sampling rate. The collected audio and video data are semantically segmented to divide the audio and video data into multiple sub-data with different semantic labels. These semantic labels may include the identity of the speaker, different parts of the conference topic, key discussion points, etc. The audio and video data are segmented using the pre-generated material distribution list and the semantic labels corresponding to the multiple groups of segmented sub-audio and video data. According to the content corresponding to each semantic label, the audio and video data is segmented into multiple groups of target sub-audio and video data. Each group of target sub-audio and video data contains specific semantic information, such as a speech segment of a specific speaker or a concentrated discussion part of a discussion topic. Transmission data analysis is performed on each group of target sub-audio and video data. Based on parameters such as resolution, frame rate, bit rate, and audio sampling rate of each group of data, its specific requirements and performance in network transmission are evaluated. The result of the transmission data analysis is the transmission parameter data corresponding to each group of target sub-audio and video data. These parameter data include information such as bandwidth requirements and transmission priority to ensure that each group of data can maintain the best quality and efficiency during transmission. The transmission parameter data corresponding to each group of target sub-audio and video data are merged to form a complete transmission parameter set. The transmission parameter set contains the transmission requirements and characteristics of all target sub-audio and video data. Through the transmission parameter set, the transmission method of audio and video data can be adaptively adjusted according to the real-time network status and data requirements to ensure efficient transmission and smooth playback of audio and video data during the meeting.

[0068] Step S104: input the transmission data set into a preset Seq2Seq model to extract conference data to obtain structured conference record data;

[0069] Specifically, video data is extracted from the transmission data set, and key frame data and metadata are obtained from the audio and video data. The metadata includes timestamps and speaker identification data. Key frame data represents representative or important images in the meeting, while metadata helps identify the specific time and speaker identity of these key frames in the meeting. At the same time, audio data is extracted from the transmission data set to obtain an audio data set, and the audio data set is converted into initial text data. Initial text data is a textual representation of audio content, usually achieved through speech recognition technology. However, initial text data may contain errors and redundant information, and data cleaning is required. The data cleaning process aims to remove errors, redundancy and irrelevant information to obtain target text data. Sequence conversion is performed on the target text data to generate a word vector set. The word vector set is a representation of text data in a high-dimensional space. By converting text into vectors, the data can be processed and analyzed more efficiently. While generating the word vector set, additional feature analysis is performed on the key frame data and metadata, more useful features are extracted from the key frames and metadata, and these features are associated with the word vector set. The additional feature set includes image features, time features, and speaker features related to the text data, which can provide more comprehensive information support. The word vector set and the additional feature set are fused to integrate feature data from different sources to form a comprehensive target word vector set. Through feature fusion, the semantic information of the text data is organically combined with the image data, time data, and speaker data to form a more comprehensive and accurate data representation. The target word vector set is input into the preset Seq2Seq model for conference data extraction. The Seq2Seq model is a sequence-to-sequence model based on a neural network for natural language processing tasks. By inputting the target word vector set, the Seq2Seq model can generate structured meeting record data. These structured data include the main content of the meeting, the speaker's views, and the key points of the discussion.

[0070] The target word vector set is input into the encoder for sequence analysis, and the input word vector sequence is converted into context vector representation data. The context vector captures the semantic and context information in the input sequence. By encoding the target word vector set one by one, the encoder generates a context vector representation data, which contains the semantic information of the entire input sequence. The generated context vector representation data is sequence converted to obtain context sequence data. The context sequence data is a further processing and organization of the context vector so that it can be efficiently processed by the decoder, ensuring that the context information can maintain the original semantic integrity and sequential logic when it is passed to the decoder. The context sequence data is input into the decoder for structured parameter analysis, and the context sequence data is converted into a structured parameter set. The structured parameter set includes information such as paragraphs, punctuation marks, and titles, which are the basic components of the meeting minutes. Through the analysis and processing of the decoder, the structured information in the input sequence is identified and extracted, and the semantic information is organized into an orderly and standardized text structure. The context sequence data is converted into structured data through the structured parameter set to generate the final structured meeting minutes data. Reorganize and format context sequence data according to structural parameters such as paragraphs, punctuation, and titles to form clear and logically coherent meeting records.

[0071] Step S105: construct a management strategy for the structured conference record data to obtain a conference management strategy, and use the conference management strategy to control the video conference in real time.

[0072] Specifically, the structured meeting record data is segmented into individual words or phrases to form a segmented set. The segmented set includes the key content and discussion points in the meeting, as well as the specific words and expressions of the speakers. The segmented set is subjected to behavioral pattern recognition to extract behavioral pattern data. The behavioral pattern data quantifies and analyzes the behavioral characteristics of each participant in the meeting, including indicators such as speaking frequency, speaking duration, and interaction duration. Through these indicators, the performance and participation of each participant in the meeting can be understood. For example, speaking frequency can show the number of times a participant speaks in a meeting, speaking duration can show the total time of his or her speaking, and interaction duration indicates the time the participant interacts with others in the meeting. The meeting rules are constructed through the behavioral pattern data to obtain the meeting rule data. The meeting rules are specific provisions for the behavioral norms and interaction methods during the meeting, such as the speaking time limit for each participant, the priority arrangement of the discussion, and the frequency requirements for interaction. Through these rules, the orderly progress of the meeting is ensured to avoid excessive speaking or insufficient interaction of individual participants. The meeting management strategy is generated based on the meeting rule data. The conference management strategy is a specific implementation plan for the conference rules, including specific operation steps and execution methods. For example, when a participant's speaking time exceeds the preset limit, the system can issue a prompt to remind the participant to control his speaking time; or when the interaction frequency is insufficient, the system can actively guide the participants to have more interaction and discussion. In order to assist in the implementation of the conference management strategy, conference management prompts are generated according to the strategy. These prompts are specific reminders and instructions for participants, such as "please speak briefly" or "please interact with other participants". Based on the conference management strategy and conference management prompts, the video conference is controlled in real time. During the conference, the behavior and interaction of each participant are continuously monitored, and real-time adjustments and controls are made according to the preset management strategy. For example, when it is found that a participant's speaking time is too long, the system will promptly issue a prompt to remind the participant to control his speech; when the interaction frequency is lower than expected, the system will actively guide and encourage participants to have more interaction and discussion. Through real-time control, the order and efficiency of the meeting can be effectively maintained, and the interactivity and participation of the meeting can also be improved.

[0073] In the embodiment of the present invention, the multifunctional video conferencing interaction method of the present invention realizes efficient identity authentication, personalized data distribution, stable audio and video transmission, intelligent conference record generation and real-time conference management by introducing advanced artificial intelligence algorithms, which greatly improves the intelligence level and user experience of video conferencing, effectively solves the deficiencies in the prior art, and has significant innovation and practical value.

[0074] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0075] (1) Collecting the conference theme and setting the conference parameters according to the conference theme, wherein the conference parameters include basic parameters and participant identity information;

[0076] (2) Extract the meeting time, meeting location and meeting agenda from the basic parameters;

[0077] (3) extracting the participant's identity information to obtain the participant's basic information, where the participant's basic information includes: name, email address, and position;

[0078] (4) constructing a virtual identity image based on the basic information of the participant to obtain multiple virtual identity images;

[0079] (5) Authentication is performed on each virtual identity image. When each virtual identity image passes the authentication, an identifier is constructed for each virtual identity image to obtain an identity identifier set.

[0080] Specifically, collect the meeting theme. Set the corresponding meeting parameters according to the meeting theme, including basic parameters and participant identity information. Basic parameters cover the time, location and agenda of the meeting, while participant identity information involves detailed information of each participant, such as name, email address and position. Extract specific details of the meeting from the basic parameters. Assume that the meeting time is set at a specific date and time period, the meeting location is held through an online meeting platform, and the meeting agenda includes multiple links, such as opening introductions, department reports, special discussions, summary speeches, etc. Extract information about the participant identity information. The basic information of each participant includes his or her name, email address and position. This information is usually obtained through the participant registration information collected in advance or the personnel database within the organization. Through the basic information, a detailed identity profile is established for each participant to facilitate subsequent management and interaction. In order to improve the interactivity and security of the meeting, a virtual identity image is constructed based on the basic information of each participant. Digitize the photo or other identification features of each participant to generate a corresponding virtual image. For example, by uploading the participant's personal photo or generating his or her virtual avatar through other digital methods. Virtual identity images are not only used for identity recognition in meetings, but also for enhancing interaction during meetings, such as displaying a virtual image when speaking to enhance recognition. Authentication is performed for each virtual identity image. Authentication usually uses biometric technology, such as facial recognition or fingerprint recognition, to ensure that the virtual identity image is consistent with the identity of the actual participant. Only when each virtual identity image is verified will a unique identifier be constructed for each virtual identity image. These identifiers are not only used for identity recognition during meetings, but also for post-meeting recording and tracking. For example, the identity of the participant is verified through facial recognition technology. When the system confirms that the virtual identity image matches its actual identity, the virtual identity image generates a unique identifier, which is used for identity recognition and record management throughout the meeting. The construction of an identifier set is an important step to ensure that the meeting is safe and orderly. All activities of each participant in the meeting, such as speaking, interacting, accessing materials, etc., will be recorded and managed through their unique identifiers. For example, during a meeting, when a participant speaks, the system will use the participant's identifier to record the content of his or her speech and mark his or her identity in the meeting minutes.

[0081] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0082] (1) Collecting historical behavior data based on participant identity information, where the historical behavior data includes: meeting participation records, speech data, and document access data;

[0083] (2) constructing behavior labels for the identity information of participants in historical behavior data to obtain multiple groups of behavior labels;

[0084] (3) constructing behavior nodes for each group of behavior labels respectively to obtain a set of behavior nodes corresponding to each group of behavior labels;

[0085] (4) Input the meeting parameters and behavior node set into the graph neural network to construct the meeting behavior graph structure and obtain the target meeting behavior graph structure;

[0086] (5) Conducting behavior correlation data analysis on the target meeting behavior graph structure to obtain a behavior correlation data set;

[0087] (6) Match the meeting information with the behavior-related data set using the basic information of the participants to obtain a data distribution list.

[0088] Specifically, historical behavior data is collected based on the participant's identity information, including meeting participation records, speech data, and document access data. For example, a participant's attendance in the past few meetings (such as how many meetings he attended, the specific time and duration of each meeting), speech data (including the number of speeches, the duration of each speech, the content summary of the speech, etc.), and document access data (for example, which meeting materials were accessed, the access frequency, and the duration of each visit, etc.). The historical behavior data is analyzed, and behavior tags are constructed based on the participant's identity information. For example, by analyzing a participant's meeting participation records, multiple behavior tags are generated, such as "high-frequency participant", "active speaker", or "deep reader of documents". These tags can reflect the behavioral characteristics and habits of participants in meetings. Behavior nodes are constructed for each group of behavior tags. Behavior nodes are the specific manifestation of behavior tags, and each node represents the specific behavior performance of a participant in a certain aspect. For example, for the "high-frequency participant" tag, the behavior node may include the specific number of meeting participations and participation time; for the "active speaker" tag, the behavior node may include the specific number of speeches and speech duration. By constructing nodes for each group of behavior labels, we can obtain the behavior node sets corresponding to each group of behavior labels. These node sets are the basic data for graph neural network analysis. The conference parameters and behavior node sets are input into the preset graph neural network model to construct the conference behavior graph structure. The graph neural network uses the relationship and connection method between nodes to construct the target conference behavior graph structure. The graph structure can show the complex relationship between participants and the content of the conference. The system can fully understand the behavior patterns and interaction relationships of each participant in the conference. For example, a behavior node may represent the speaking behavior of a participant in a conference, while another behavior node represents the data access behavior of the participant in the same conference. Through the graph neural network, the relationship between these nodes can be revealed. Perform behavior association data analysis on the target conference behavior graph structure. Identify and mine the correlation between behavior nodes to form a behavior association data set. The association data can reveal the behavior trends and interaction patterns of participants in the conference. Based on the basic information of the participants, the behavior association data set is matched with the conference data to generate the final data distribution list. For example, if a participant has shown a high level of interest in a particular topic and has spoken frequently in past meetings, the system will prioritize distributing materials related to that topic to that participant in new meetings. This personalized material distribution method improves the efficiency of meeting material utilization and the satisfaction of participants.

[0089] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0090] (1) Collect audio and video data in real time, analyze the parameters of the audio and video data, and obtain the resolution, frame rate, bit rate, and audio sampling rate;

[0091] (2) performing semantic segmentation on the audio and video data to obtain semantic labels corresponding to multiple groups of sub-audio and video data;

[0092] (3) segmenting the audio and video data using the data distribution list and the semantic labels corresponding to the multiple groups of sub-audio and video data to obtain multiple groups of target sub-audio and video data;

[0093] (4) Based on the resolution, frame rate, bit rate and audio sampling rate, each group of target sub-audio and video data is analyzed for transmission data, transmission parameter data corresponding to each group of target sub-audio and video data is obtained, and the transmission parameter data corresponding to each group of target sub-audio and video data is merged into a transmission parameter set.

[0094] Specifically, audio and video data are collected in real time, including video streams and audio streams in the meeting. Parameter analysis is performed on the audio and video data to obtain its technical parameters, such as video resolution, frame rate, bit rate, and audio sampling rate. Semantic segmentation is performed on the audio and video data to divide the audio and video data into multiple sub-data with different semantic labels. For example, different clips in the video may correspond to different meeting sessions, such as speeches, discussions, demonstrations, etc., while different clips in the audio data may correspond to different speech content or speakers. Through semantic segmentation, each group of sub-audio and video data is assigned a corresponding semantic label to help the system identify and process different types of data in subsequent processing. The audio and video data is segmented by using a pre-generated data distribution list and semantic labels corresponding to multiple groups of sub-audio and video data. The data distribution list usually contains detailed information about the participants and their data needs and preferences, while the semantic labels help the system identify the specific content and purpose of each group of sub-audio and video data. For example, if a participant has shown a high interest in a specific topic in a past meeting, the relevant audio and video data is preferentially segmented according to the participant's data distribution list and semantic labels to form multiple groups of target sub-audio and video data. Based on the resolution, frame rate, bit rate and audio sampling rate, each group of data is analyzed for transmission data, the specific requirements and performance of each group of data in network transmission are evaluated, and appropriate transmission resources are allocated to it. For example, for video data with higher resolution and higher frame rate, more bandwidth and higher transmission priority are allocated to it to ensure that it maintains high quality and smoothness during transmission. Correspondingly, for audio data, its transmission requirements are evaluated according to its sampling rate and bit rate to ensure the clarity and synchronization of the audio. The transmission parameter data corresponding to each group of target sub-audio and video data are merged to form a complete transmission parameter set. The transmission parameter set contains the transmission requirements and characteristics of all target sub-audio and video data. Through the transmission parameter set, the transmission method of audio and video data is adaptively adjusted according to the real-time network status and data requirements to ensure efficient transmission and smooth playback of audio and video data during the meeting.

[0095] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0096] (1) extracting video data from the transmission data set to obtain key frame data and metadata, wherein the metadata includes: timestamp and speaker identification data;

[0097] (2) extracting audio data from the transmission data set to obtain an audio data set, and converting the audio data set into initial text data;

[0098] (3) Perform data cleaning on the initial text data to obtain the target text data;

[0099] (4) Perform sequence conversion on the target text data to obtain a word vector set;

[0100] (5) performing additional feature analysis on the word vector set, key frame data, and metadata to obtain an additional feature set corresponding to the word vector set;

[0101] (6) Perform feature fusion on the word vector set and the additional feature set to obtain the target word vector set;

[0102] (7) Input the target word vector set into the preset Seq2Seq model to extract meeting data and obtain structured meeting record data.

[0103] Specifically, video data is extracted from the transmission data set, and key frame data is identified and captured. Key frames refer to frames in the video stream that can represent important content, such as speaker switching, important presentation content, etc. Metadata is extracted, and the metadata includes the timestamp and speaker identification data of each key frame. The timestamp is used to mark the specific time point of each key frame in the video stream, and the speaker identification data is used to identify the specific person who spoke at that time point. Audio data is extracted from the transmission data set to obtain an audio data set. The audio data set contains all the voice information in the meeting, and the audio data is converted into initial text data. The conversion process usually uses automatic speech recognition technology to convert the voice signal into the corresponding text record. The initial text data may contain some misidentification and noise, and data cleaning is required to remove errors, redundancy and irrelevant information to obtain the target text data. The target text data is sequence converted to generate a word vector set. The word vector set is a vector representation of text data converted into a high-dimensional space. In this way, the semantic information of the text data can be processed by the system in a digital form. Common word vector generation methods include Word2Vec, GloVe, and BERT, which can capture semantic relationships and contextual information in the text. Additional feature analysis is performed on the word vector set, key frame data, and metadata to extract more useful features from the key frames and metadata and associate these features with the word vector set. The additional feature set includes image features, time features, and speaker features related to text data. For example, the image content in the key frame is analyzed to extract important visual features and associate these features with the corresponding text data; at the same time, the speaker's speech content and behavior characteristics at different time points are analyzed using timestamp information to form a comprehensive feature set. Feature fusion is performed on the word vector set and the additional feature set to integrate feature data from different sources to form a comprehensive target word vector set. Through feature fusion, the semantic information of the text data is combined with the image data, time data, and speaker data to form a more comprehensive and accurate data representation. The target word vector set is input into the preset Seq2Seq model for conference data extraction. The Seq2Seq model is a sequence-to-sequence model based on neural networks and is widely used in natural language processing tasks. By inputting the target word vector set, the Seq2Seq model can generate structured conference record data. These structured data include the main content of the meeting, the speakers’ opinions and the key points of the discussion, providing a clear and systematic text representation for meeting records and subsequent analysis.

[0104] In a specific embodiment, the process of executing the step of inputting the target word vector set into a preset Seq2Seq model to extract conference data may specifically include the following steps:

[0105] (1) Input the target word vector set into the encoder for sequence analysis to obtain the context vector representation data;

[0106] (2) performing sequence conversion on the context vector representation data to obtain context sequence data;

[0107] (3) Inputting the context sequence data into the decoder for structured parameter analysis to obtain a structured parameter set, wherein the structured parameter set includes paragraphs, punctuation marks, and titles;

[0108] (4) The context sequence data is converted into structured data using a structured parameter set to obtain structured meeting record data.

[0109] Specifically, the target word vector set is input into the encoder for sequence analysis, and the target word vector set is converted into context vector representation data. The target word vector set is a high-dimensional vector representation obtained after sequence conversion of the conference text data. Through the processing of the encoder, the contextual relationship between words and the overall semantic structure are captured to generate context vector representation data. The context vector representation data is sequence converted to generate context sequence data. The purpose of sequence conversion is to organize the context vector representation data in a certain order to facilitate the processing of the decoder. For example, in the meeting minutes, the speeches of different speakers need to be arranged in chronological order to ensure the logic and readability of the records. The context sequence data is input into the decoder for structured parameter analysis. The context sequence data is converted into a structured parameter set, which includes information such as paragraphs, punctuation marks, and titles. Through the analysis of the decoder, the structured information in the text is identified, such as the paragraph division of the speech content, the punctuation marks between sentences, and the titles of each part. The context sequence data is converted into structured data through the structured parameter set, and the context sequence data is reorganized and typeset according to the structured parameters to generate the final structured meeting record data. For example, the content of different speakers is divided into multiple paragraphs according to the paragraph information, punctuation is used to separate sentences, and appropriate titles are added to identify different discussion topics or meeting sessions. The context sequence data is input into the decoder for structured parameter analysis. The decoder identifies the paragraphs, punctuation, and title information in the text. For example, the speech at the beginning of the meeting, the report content of each department, the different parts of the special discussion, and the final summary speech are identified. Through structured parameters, the system can accurately divide the text content and add appropriate identifiers. The context sequence data is converted into structured data through the structured parameter set to generate structured meeting record data. Different speech contents are divided into multiple paragraphs according to the paragraph information, and punctuation is used to separate sentences.

[0110] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0111] (1) Perform text segmentation on structured conference record data to obtain a segmentation set;

[0112] (2) performing behavioral pattern recognition on the word set to obtain behavioral pattern data, wherein the behavioral pattern data includes: speaking frequency, speaking duration, and interaction duration;

[0113] (3) Constructing meeting rules through behavioral pattern data to obtain meeting rule data;

[0114] (4) Generate a conference management strategy through conference rule data, and generate conference management prompt words according to the conference management strategy;

[0115] (5) Real-time control of video conferences based on conference management strategies and conference management prompts.

[0116] Specifically, the structured conference record data is segmented to obtain a segmentation set. Text segmentation is one of the basic steps in natural language processing, which divides continuous text into separate words or phrases for subsequent analysis. The segmentation technology can use a dictionary-based method or a more complex machine learning method, such as a segmenter and a pre-trained language model. Through text segmentation, each sentence in the structured conference record data is broken down into basic vocabulary units to form a segmentation set. Behavioral pattern recognition is performed on the segmentation set, and behavioral data related to the conference participants are extracted from the segmentation set. These behavioral data include speaking frequency, speaking duration, and interaction duration. These data can be obtained by analyzing the timestamps and participant identity tags in the segmentation set. For example, by counting the number of speeches of each participant and the duration of each speech, the speaking frequency and speaking duration are calculated. By analyzing the frequency and duration of interactive dialogues, the interaction duration data is obtained. The meeting rules are constructed based on the behavioral pattern data. Meeting rules are an important basis for meeting management, which aims to standardize the conduct of meetings and ensure the orderliness and efficiency of meetings. For example, by analyzing the behavior pattern data, if it is found that some participants speak too frequently or speak for too long, it may reduce the speaking opportunities of other participants and affect the overall efficiency of the meeting. Therefore, the meeting rules are constructed, such as the speaking time limit and the speaking frequency limit for each participant. In addition, according to the interaction time data, the minimum interaction time requirement is set to promote interaction and discussion between participants. The meeting management strategy is generated through the meeting rule data. The meeting management strategy is a specific implementation plan, including how to apply and execute these rules in the actual meeting process. For example, some automated management measures are set, such as when a participant's speaking time exceeds the preset limit, the system automatically issues a prompt; when the interaction frequency is insufficient, the system automatically reminds the host to guide the discussion, etc. In order to assist in the implementation of the meeting management strategy, the meeting management prompt words are generated. These prompt words are specific reminders and guidance for participants, such as "please speak briefly" and "please interact with other participants". The prompt words can be dynamically adjusted according to the real-time meeting situation and the preset management strategy, so as to achieve effective management of the meeting. Based on the meeting management strategy and the meeting management prompt words, the video conference is controlled in real time. During the meeting, the system continuously monitors the behavior and interaction of each participant, and makes real-time adjustments and controls based on the preset management strategies. For example, if a participant speaks for too long, the system will promptly issue a prompt to remind the participant to control his or her speaking time; when the interaction frequency is lower than expected, the system will actively guide and encourage participants to have more interactions and discussions.

[0117] The above describes the multifunctional video conferencing interaction method in the embodiment of the present application. The following describes the multifunctional video conferencing interaction device in the embodiment of the present application. Figure 2In the embodiment of the present application, an embodiment of the multifunctional video conference interaction device includes:

[0118] The collection module 201 is used to collect the conference theme and set the conference parameters according to the conference theme, wherein the conference parameters include: basic parameters and participant identity information, and perform identity authentication on the conference parameters to obtain an identity identifier set;

[0119] An input module 202 is used to collect historical behavior data based on the participant identity information, and input the conference parameters and the historical behavior data into a preset graph neural network to perform conference data analysis to obtain a data distribution list;

[0120] The analysis module 203 is used to collect audio and video data in real time, and perform adaptive transmission data analysis on the audio and video data according to the data distribution list to obtain a transmission parameter set;

[0121] An extraction module 204 is used to input the transmission data set into a preset Seq2Seq model to extract conference data to obtain structured conference record data;

[0122] The construction module 205 is used to construct a management strategy for the structured conference record data to obtain a conference management strategy, and to control the video conference in real time through the conference management strategy.

[0123] Through the collaborative cooperation of the above-mentioned components, the present invention introduces a Seq2Seq model based on an attention mechanism for intelligent generation of meeting records. The model can automatically convert the speech content in the meeting into structured meeting records, greatly reducing the workload and error rate of manual recording. The Seq2Seq model processes long segments of speech through an encoder-decoder structure and focuses on key content through an attention mechanism to generate accurate and coherent meeting records. This function not only improves the accuracy and completeness of the meeting records, but also facilitates the review and follow-up of the meeting content after the meeting. The present invention constructs an intelligent meeting management strategy through behavioral pattern recognition and text analysis. The system can monitor the progress of the meeting in real time, analyze the frequency of speaking, the duration of speaking and the interactive situation, and generate real-time feedback reports and optimized management strategies. These strategies include speech order adjustment, time management and interactive control, etc., to ensure that the meeting is carried out in an orderly manner and improve the efficiency and effectiveness of the meeting. Through real-time monitoring and feedback, the system can dynamically adjust the meeting schedule and solve various problems that may arise during the meeting. In addition, the present invention can also store the generated structured meeting record data in a secure database, and perform access control through identity identifiers to ensure the security and privacy of meeting materials.

[0124] The present application also provides a multifunctional video conferencing interaction device, which includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the multifunctional video conferencing interaction method in the above-mentioned embodiments.

[0125] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the multi-functional video conferencing interaction method.

[0126] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0127] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially 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, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0128] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multifunctional video conference interaction method, characterized in that: The multifunctional video conference interaction method comprises: Collecting the conference theme and setting the conference parameters according to the conference theme, wherein the conference parameters include: basic parameters and participant identity information, and performing identity authentication on the conference parameters to obtain an identity identifier set; Based on the participant identity information, historical behavior data is collected, and the conference parameters and the historical behavior data are input into a preset graph neural network for conference data analysis to obtain a data distribution list; specifically including: based on the participant identity information, historical behavior data is collected, wherein the historical behavior data includes: conference participation records, speech data, and data access data; behavior labels are constructed for the participant identity information according to the historical behavior data to obtain multiple groups of behavior labels; behavior nodes are constructed for each group of the behavior labels respectively to obtain a behavior node set corresponding to each group of the behavior labels; the conference parameters and the behavior node set are input into the graph neural network to construct a conference behavior graph structure to obtain a target conference behavior graph structure; behavior association data analysis is performed on the target conference behavior graph structure to obtain a behavior association data set; meeting data is matched with the behavior association data set through the participant basic information to obtain the data distribution list; Collect audio and video data in real time, and perform adaptive transmission data analysis on the audio and video data according to the data distribution list to obtain a transmission data set; Inputting the transmission data set into a preset Seq2Seq model to extract conference data to obtain structured conference record data; A management strategy is constructed for the structured conference record data to obtain a conference management strategy, and the video conference is controlled in real time through the conference management strategy.

2. The multifunctional video conference interaction method according to claim 1, characterized in that: The collecting of conference topics and setting of conference parameters according to the conference topics, wherein the conference parameters include: basic parameters and participant identity information, and identity authentication of the conference parameters to obtain an identity identifier set, including: Collecting the conference theme and setting the conference parameters according to the conference theme, wherein the conference parameters include: basic parameters and participant identity information; Extracting the meeting time, meeting location and meeting agenda from the basic parameters; Extracting the participant's identity information to obtain basic information of the participant, wherein the basic information of the participant includes: name, email address and position; Constructing a virtual identity image based on the basic information of the participant to obtain multiple virtual identity images; Authentication is performed on each of the virtual identity images. When each of the virtual identity images passes the authentication, an identifier is constructed for each of the virtual identity images to obtain the identity identifier set.

3. The multifunctional video conference interaction method according to claim 1, characterized in that: The real-time collection of audio and video data, and the adaptive transmission data analysis of the audio and video data according to the data distribution list to obtain a transmission data set, include: Collect audio and video data in real time, perform parameter analysis on the audio and video data, and obtain resolution, frame rate, bit rate, and audio sampling rate; Performing semantic segmentation on the audio and video data to obtain semantic labels corresponding to multiple groups of sub-audio and video data; Segmenting the audio and video data using the data distribution list and the semantic tags corresponding to the multiple groups of sub-audio and video data to obtain multiple groups of target sub-audio and video data; Based on the resolution, frame rate, bit rate and audio sampling rate, transmission data analysis is performed on each group of target sub-audio and video data to obtain transmission parameter data corresponding to each group of target sub-audio and video data, and the transmission parameter data corresponding to each group of target sub-audio and video data are merged into the transmission data set.

4. The multifunctional video conference interaction method according to claim 1, characterized in that: The step of inputting the transmission data set into a preset Seq2Seq model to extract conference data to obtain structured conference record data includes: Extracting video data from the transmission data set to obtain key frame data and metadata, wherein the metadata includes: a timestamp and speaker identification data; Extracting audio data from the transmission data set to obtain an audio data set, and converting the audio data set into initial text data; Performing data cleaning on the initial text data to obtain target text data; Performing sequence conversion on the target text data to obtain a word vector set; Performing additional feature analysis on the word vector set, the key frame data, and the metadata to obtain an additional feature set corresponding to the word vector set; Performing feature fusion on the word vector set and the additional feature set to obtain a target word vector set; The target word vector set is input into a preset Seq2Seq model to extract conference data to obtain structured conference record data.

5. The multifunctional video conference interaction method according to claim 4, characterized in that: The Seq2Seq model includes an encoder and a decoder. The target word vector set is input into a preset Seq2Seq model to extract conference data to obtain structured conference record data, including: Inputting the target word vector set into the encoder for sequence analysis to obtain context vector representation data; Performing sequence conversion on the context vector representation data to obtain context sequence data; Inputting the context sequence data into the decoder for structured parameter analysis to obtain a structured parameter set, wherein the structured parameter set includes: paragraphs, punctuation marks, and titles; The context sequence data is converted into structured data using the structured parameter set to obtain the structured meeting record data.

6. The multifunctional video conference interaction method according to claim 1, characterized in that: The step of constructing a management strategy for the structured conference record data to obtain a conference management strategy, and controlling the video conference in real time by using the conference management strategy, includes: Performing text segmentation on the structured conference record data to obtain a segmentation set; Performing behavior pattern recognition on the word segmentation set to obtain behavior pattern data, wherein the behavior pattern data includes: speech frequency, speech duration, and interaction duration; Constructing conference rules through the behavior pattern data to obtain conference rule data; Generate a conference management strategy through the conference rule data, and generate a conference management prompt word according to the conference management strategy; The video conference is controlled in real time based on the conference management strategy and the conference management prompt words.

7. A multifunctional video conference interactive device, characterized in that: The multifunctional video conference interactive device comprises: A collection module, used to collect conference topics and set conference parameters according to the conference topics, wherein the conference parameters include: basic parameters and participant identity information, and perform identity authentication on the conference parameters to obtain an identity identifier set; An input module is used to collect historical behavior data based on the participant identity information, and input the conference parameters and the historical behavior data into a preset graph neural network to perform conference data analysis to obtain a data distribution list; specifically comprising: collecting historical behavior data based on the participant identity information, wherein the historical behavior data includes: conference participation records, speech data, and data access data; constructing behavior labels for the participant identity information based on the historical behavior data to obtain multiple groups of behavior labels; constructing behavior nodes for each group of the behavior labels respectively to obtain a behavior node set corresponding to each group of the behavior labels; inputting the conference parameters and the behavior node set into the graph neural network to construct a conference behavior graph structure to obtain a target conference behavior graph structure; performing behavior association data analysis on the target conference behavior graph structure to obtain a behavior association data set; matching the behavior association data set with the participant basic information to obtain the data distribution list; An analysis module, used for collecting audio and video data in real time, and performing adaptive transmission data analysis on the audio and video data according to the data distribution list to obtain a transmission data set; An extraction module, used for inputting the transmission data set into a preset Seq2Seq model to extract conference data and obtain structured conference record data; The construction module is used to construct a management strategy for the structured conference record data, obtain a conference management strategy, and control the video conference in real time through the conference management strategy.

8. A multifunctional video conference interactive device, characterized in that: The multifunctional video conference interactive device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the multifunctional video conference interaction device to execute the multifunctional video conference interaction method according to any one of claims 1 to 6.

9. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the multifunctional video conferencing interaction method as described in any one of claims 1 to 6 is implemented.

Images