An intelligent and digital fusion communication method and system based on artificial intelligence

Abstract

The application discloses a kind of digital intelligence fusion communication method and system based on artificial intelligence, and SIP application module is the conference priority of each conference that edge media server carries;Digital intelligence control module obtains the low priority conference of the resource occupancy density of pre-set quantity, and the low priority conference of few number of participants;Digital intelligence control module real-time collection performance parameter of edge media server;Using trained prediction model to obtain the load prediction condition of the edge media server;According to the load prediction condition, dynamically adjust media stream distribution mode.The load of edge media server is predicted in the application, according to the priority of conference, number of participants and load prediction condition, flexibly select media stream distribution mode, meet the demand of user to video communication.

Description

TECHNICAL FIELD

[0001] The application relates to the field of fusion communication, in particular to an intelligent fusion communication method and system based on artificial intelligence. BACKGROUND

[0002] In recent years, multimedia technology and data communication technology have developed rapidly. Video conferencing can solve the problem of long-distance communication, but the traditional video conference system has many shortcomings, which affects its wide application in the market. For example: 1) for general SIP video conference systems, only SIP-based client connections can be supported, and more types of clients cannot be accessed, and there are deficiencies in compatibility; 2) in terms of audio and video transmission quality, the traditional video conference system often has problems such as incoherent pictures and audio flow interruption during operation, which makes the communication between users in the system not smooth.

[0003] The fusion communication video conference system is a platform integrating multiple communication technologies, aiming to provide a comprehensive video conference solution. One of the main advantages of the fusion communication video conference system is its ease of use. Users can easily join a meeting through a web browser or mobile application without the need to install additional software. This makes video conferencing more popular and accessible.

[0004] However, the fusion communication video conference system also faces some challenges, such as high signaling delay, insufficient compatibility, bandwidth limitations, and other issues in large-scale cross-protocol sessions. To address these challenges, most systems rely on static rules such as fixed bandwidth thresholds or "one-size-fits-all" global resolution reduction, lacking predictive optimization for complex network environments, resulting in severe resource contention and reduced service quality during burst traffic. SUMMARY

[0005] The purpose of the present application is to provide an intelligent fusion communication method and system based on artificial intelligence, which can predict the load of the edge media server, select the media stream distribution mode flexibly according to the priority of the conference, the number of participants and the load prediction, and meet the user's demand for video communication.

[0006] The technical solution adopted by the application to solve its technical problems is as follows:

[0007] In a first aspect, an intelligent fusion communication method based on artificial intelligence is provided. A SIP application module sets a conference priority for each conference carried by an edge media server. The conference client includes a SIP terminal and / or a WebRTC terminal. The SIP terminal is directly connected to the conference server, and the WebRTC terminal is connected to the conference server via a WebRTC server and a gateway. The conference server includes a SIP application module and an intelligent control module.

[0008] The digital control module sorts the low-priority meetings according to the resource occupation density from high to low, and obtains the first N low-priority meetings with high resource occupation density; the digital control module sorts the low-priority meetings according to the number of participants from low to high, and obtains the first N low-priority meetings with few participants; the value of N is preset;

[0009] The digital control module collects the performance parameters of the edge media server in real time; the trained prediction model is used to obtain the load prediction of the edge media server; and the media stream distribution mode is dynamically adjusted according to the load prediction, specifically including:

[0010] When the load prediction is too large, for the first N low-priority meetings with high resource occupation density, the central media server takes over the media stream of the meeting; for the first N low-priority meetings with few participants, the media stream distribution mode of the meeting is adjusted to a distributed media stream transmission mode.

[0011] Further, the prediction model includes a preprocessing module, an LSTM layer, an attention mechanism layer, a full connection layer, an activation function, and an output layer; the preprocessing module aggregates, denoises, and normalizes the training set data according to a time window to generate a feature matrix; the feature matrix is input into the LSTM layer to capture short-term local time series fluctuations; the attention mechanism layer is used to identify key time points; the full connection layer is used for global feature integration to output load prediction for a future preset time length; the activation function is used to capture the complex interaction between the performance parameters and enhance the non-linear expression ability; and the output layer is used to directly output the load value for measuring the load size.

[0012] Further, the resource occupation density is the number of participants divided by the meeting occupation bandwidth.

[0013] Further, for the first N low-priority meetings with high resource occupation density, the central media server takes over the media stream of the meeting, including: a SIP application module sends a notification message to the central media server and each terminal in the meeting, the central media server and each terminal in the meeting returns an acknowledgement message after receiving the notification message, and each terminal in the meeting modifies the sending address and port and the receiving address and port of the terminal media stream according to the parameter information of the notification message, so that the central media server takes over the media stream of the meeting.

[0014] Further, the low-priority conference with less than N participants is adjusted to a media stream distribution mode, including: the SIP application module sends a notification message to each terminal in the conference, each terminal in the conference returns an acknowledgement message after receiving the notification message, and modifies the sending address and port and the receiving address and port of the terminal media stream according to the parameter information of the notification message, so as to change the centralized media stream transmission mode to a distributed media stream transmission mode.

[0015] Further, the performance parameters of the edge media server include CPU usage, memory occupancy, current call number, codec queue length and bandwidth utilization.

[0016] Further, the SIP application module sets the conference priority for each conference carried by the edge media server, including: the SIP application module sets the conference priority according to the client identifier participating in the conference, and when there is a specific client participating, the SIP application module automatically sets the conference as high priority.

[0017] In a second aspect, a digital and intelligent converged communication system based on artificial intelligence is provided, including: a central service platform and a plurality of domains; wherein the central service platform includes a location server and a central media server; the domains include a conference server, a WebRTC server, a gateway, a client and an edge media server; the client includes a SIP client and a WebRTC client;

[0018] The conference server includes a SIP application module and a digital and intelligent control module;

[0019] The gateway is located between the conference server and the WebRTC server, and is used to convert WebRTC signaling and SIP signaling to each other, and complete mutual communication between heterogeneous networks;

[0020] The SIP terminal is directly connected to the conference server, and the WebRTC terminal is connected to the conference server via the WebRTC server and the gateway;

[0021] The location server is used to record and maintain the domain information of the global user;

[0022] The central media server is used to take over part of the media stream of the edge media server when the load of the edge media server in each domain is too large.

[0023] Traditional video conference systems have problems such as insufficient compatibility and poor audio and video transmission quality. In the present application, the load of the edge media server is predicted through a digital control module, and the media stream distribution mode is dynamically adjusted to ensure that the media server will not cause the audio and video transmission quality to decrease due to excessive load when processing media streams, and to ensure the stability and quality of the conference, thereby providing users with a good video conference experience. Moreover, the system can flexibly select the media stream distribution mode according to the priority of the conference, the number of participants and the load prediction. For important high-priority conferences, the centralized transmission mode can be continued to be used, and the edge media server or the center media server can be used to ensure the stability and high quality of the conference. For low-priority conferences with special resource occupation, such as conferences with high resource occupation density or a small number of participants, the center media server can be used to take over or adjust to the distributed transmission mode, thereby meeting the diversified needs of different conferences, avoiding problems such as lag and crash of the edge media server due to resource exhaustion, and ensuring the stability of the entire system.

[0024] The technical solutions of the present application will be described in further detail below with reference to the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0025] The above and other objects, features and advantages of the present application will become more apparent from the following description of the embodiments of the present application with reference to the accompanying drawings, in which:

[0026] Figure 1 is a centralized media stream transmission structure diagram.

[0027] Figure 2 is a distributed media stream transmission structure diagram.

[0028] Figure 3 is a structure diagram of a converged communication system provided by an exemplary embodiment of the present application.

[0029] Figure 4 is an intra-domain structure diagram of a converged communication system provided by an exemplary embodiment of the present application.

[0030] Figure 5 is a flowchart of a method provided by an exemplary embodiment of the present application.

[0031] Figure 6 is a prediction model structure diagram provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0032] The present application is described in detail below based on examples, but the present application is not limited to these examples. In the following detailed description of the present application, some specific details are described in detail. The present application can also be fully understood without the description of these details. In order to avoid confusion of the essence of the present application, the well-known methods, processes, procedures, elements and circuits are not described in detail.

[0033] In addition, those skilled in the art should understand that the drawings provided herein are for illustrative purposes only, and the drawings are not necessarily drawn to scale.

[0034] SIP is a protocol for multimedia session control, mainly used to establish, modify and terminate multimedia sessions. It was originally designed for IP telephony, but its functionality goes far beyond that. SIP supports the transmission of video and audio, so it can be used for video conferencing. SIP relies on RTP (Real-time Transport Protocol) to transmit actual media data when transmitting data. The main advantage of SIP is its flexibility and scalability, which can communicate through various network topologies and support multiple media types.

[0035] The transmission of media streams between clients includes centralized and distributed. As Figure 1 The centralized media stream transmission structure diagram includes a media server MCU that centrally processes media streams and client nodes. The audio and video streams are received, coded, mixed by the media server MCU, and then sent to each client node. For any client in the network, only a single stream needs to be transmitted, and other tasks are handled by the MCU, greatly saving client resources and improving system response speed. However, when more clients enter the conference, the media server will integrate and forward more media streams, consuming a large amount of media server resources, so the media server carrying capacity will become the bottleneck of network expansion.

[0036] As Figure 2 The distributed media stream transmission structure diagram does not have a media processor MCU between the client nodes, and the media stream is directly transmitted by the participating client nodes. Each client in the conference has a connection relationship. The advantage of this network is that media distribution and management do not need a media server between clients to achieve multi-party communication. However, the disadvantage is that it cannot provide communication services for large conference systems. The audio and video transcoding and mixing of the system are completed at the terminal, and with the increase in the number of participants in the network, the CPU, memory and bandwidth resources of the terminal will be consumed.

[0037] In recent years, multimedia technology and data communication technology have developed rapidly. Video conferencing can solve the problem of long-distance communication, but the traditional video conferencing system has many shortcomings, which affects its wide application in the market. For example: 1) For general SIP video conferencing systems, only SIP-based client connections can be supported, and more types of clients cannot be accessed, and there are deficiencies in compatibility; 2) In terms of audio and video transmission quality, the traditional video conferencing system often has problems such as discontinuous pictures and audio flow during operation, making communication between users in the system difficult.

[0038] WebRTC is a technology that supports real-time communication between web browsers. It allows real-time transmission of audio, video and data between web browsers without additional plugins or software. The core components of WebRTC include PeerConnection API, MediaStream API and DataChannel API. PeerConnection API is used to establish and manage peer connections, MediaStream API is used to handle audio and video streams, and DataChannel API is used to transmit data.

[0039] The converged communication conference system in the present application provides a more flexible and powerful video conferencing solution by combining the advantages of SIP and WebRTC. SIP can be used for session control, while WebRTC can be used for real-time media transmission. This combination enables the system to support a variety of devices and platforms, including PCs, mobile devices and embedded systems.

[0040] As shown in Figures 3-4 The converged communication conference system based on artificial intelligence in the present application includes a central service platform and multiple domains. The central service platform includes a location server and a central media server. The domains include a conference server, a WebRTC server, a gateway, a client and an edge media server. The client includes a SIP client and a WebRTC client.

[0041] The conference server includes a SIP application module and a digitalization control module. The SIP application module is mainly used for: (1) conference control, completing centralized processing of client signaling; (2) responsible for analyzing data from each client and storing client session information in the database, providing data for subsequent dynamic adjustment of media stream distribution mode; (3) setting conference priority for each conference carried by the edge media server.

[0042] The client carries a conference priority identifier in a request when initiating a conference, and the SIP application module receives the client request and sets the priority of the conference according to the priority identifier. For example, the client can set a high priority for an important conference and a low priority for an ordinary conference. Alternatively, the SIP application module can set the priority of the conference according to the identifier of the client participating in the conference, and automatically set the conference as a high priority when certain specific clients participate.

[0043] The digital control module is used to predict the load of the edge media server in the same domain, and dynamically adjust the media stream distribution mode of each client according to the load prediction. The specific prediction and adjustment mode will be described in detail below.

[0044] Since WebRTC and SIP use different solutions for conference control, WebRTC uses the JSEP protocol, and the SIP application module cannot directly manage the WebRTC client. Therefore, a separate WebRTC server is needed in the system to perform control work, and the interaction of each client is completed at the signaling layer.

[0045] In order to meet the system compatibility, the conversion between SIP and WebRTC is fully completed in the system, and the communication between SIP clients and WebRTC clients is realized. The system described in the present application adds a gateway between the conference server and the WebRTC server, which is used to convert WebRTC signaling and SIP signaling to each other and complete the mutual communication between heterogeneous networks. The signaling conversion module in the gateway is used to realize the conversion of multiple signaling protocols. When receiving the signaling transmitted by the client, the signaling conversion module first analyzes and extracts the signaling message, and then converts it into a specified intermediate signaling, which is transmitted to the WebRTC server and the conference server, and further transmitted to each client.

[0046] The client group includes SIP terminals and WebRTC terminals. The SIP terminal is directly connected to the conference server, and the WebRTC terminal needs to pass through the gateway to convert the information before uploading to the conference server.

[0047] When multiple domains are deployed, the problem is how to determine the target user's belonging domain, at this time, the location server needs to be matched for addressing, the location server records and maintains the global user's domain information, and it only manages the task of informing the target user's belonging domain to the calling SIP application module, and the subsequent service is processed by the SIP application module of the related domain. The message forwarding between domains is uniformly processed by the SIP application module of each domain, and the calling domain will not directly access the database of other domains. The SIP application module maintains the user list of the domain, and when the call object is not the user of the domain, the SIP application module of the calling user sends a request message to the location server, and the location server queries the domain information of the called user, and describes the target SIP application module of the called in the returned message. The calling SIP application module forwards the message to the target SIP application module.

[0048] The central media server is used to take over the media stream that the edge media server in each domain is unable to support. When the edge media server of a certain domain is overloaded, it will transfer part of the media stream of the client to the central media server for takeover under the decision of the intelligent control module.

[0049] The system execution process is as follows: the SIP client or WebRTC client initiates a conference request to the server, a video conference group is created in the system, and other users are invited through the conference server, or the user actively applies to join after obtaining the conference group information. The SIP client is directly controlled by the conference server during the video conference process, and the WebRTC client is controlled by the conference server through the gateway. In the media stream control, the intelligent control module is used for load prediction of the edge media server in the same domain, and according to the load prediction, the media stream distribution mode of each client is dynamically adjusted.

[0050] Embodiment 1: The present application provides an intelligent fusion communication method based on artificial intelligence, as shown in Figure 5 The SIP application module sets the conference priority for each conference borne by the edge media server; the conference client includes a SIP terminal and / or a WebRTC terminal; the SIP terminal is directly connected with the conference server, and the WebRTC terminal is connected to the conference server via a WebRTC server and a gateway; the conference server includes a SIP application module and an intelligent control module;

[0051] The intelligent control module sorts the low-priority conferences according to the resource occupation density from high to low, and obtains the first N low-priority conferences with high resource occupation density; the intelligent control module sorts the low-priority conferences according to the number of participants from low to high, and obtains the first N low-priority conferences with few participants; the value of N is preset;

[0052] The digital and intelligent control module collects performance parameters of the edge media server in real time; a trained prediction model is used to obtain a load prediction of the edge media server; and a media stream distribution mode is dynamically adjusted according to the load prediction, specifically including:

[0053] When the load prediction is too large, the media stream of the first N low-priority conferences with high resource occupation density is taken over by the central media server, and the media stream distribution mode of the first N low-priority conferences with a small number of participants is adjusted to a distributed media stream transmission mode.

[0054] When the load prediction is normal, the conference media stream taken over by the central media server is returned to the edge media server, and the media stream distribution mode of the conference using the distributed media stream transmission mode is adjusted to a centralized media stream transmission mode.

[0055] By default, a centralized media stream transmission mode is used. Since a conference is usually held among friends or staff, most participants are located within the same domain, and a small number of participants are located in different domains. Therefore, the SIP application module receives a conference request sent by a client initiating a conference and arranges an edge media server in the same domain to provide media stream transmission services for the conference.

[0056] The current number of active call sessions refers to the number of independent media stream channels that the media server is currently processing in parallel. One conference may occupy multiple call sessions, for example, one person may occupy 1-3 call sessions (video stream, audio stream, screen sharing stream). If a conference with 10 participants has one main speaker who performs screen sharing and opens a video, the media server needs to process 10 audio streams + 1 video stream + 1 screen sharing stream.

[0057] The codec queue length refers to the number of audio and video data packets waiting to be processed in the media server. The encoding queue refers to the queue of raw media streams (such as YUV frames captured by a camera) uploaded by the client waiting to be compressed (such as converted to H.264). The decoding queue refers to the queue of compressed streams (such as H.264 packets from other terminals) received waiting to be decompressed (restored to YUV / PCM). The codec queue length is usually measured in the number of data packets or frames (such as 15 frames waiting to be processed in the queue).

[0058] The resource occupation density = number of participants / conference bandwidth. For example, conference A: 10 people, 50 Mbps bandwidth, then the resource occupation density of this conference = 0.2 people / Mbps. Conference B: 20 people, 200 Mbps bandwidth, then the resource occupation density of this conference = 0.1 people / Mbps. Conference A is a conference with high resource occupation density.

[0059] For the first N low-priority conferences with high resource occupation density, the central media server takes over the media stream of the conference, including: the SIP application module sends a notification message to the central media server and each terminal in the conference, the central media server and each terminal in the conference returns an acknowledgement message after receiving the notification message, each terminal in the conference modifies the sending address and port and the receiving address and port of the terminal media stream according to the parameter information of the notification message, the central media server takes over the media stream of the conference, and the edge MCU resource is released.

[0060] For the first N low-priority conferences with few participants, the media stream distribution mode of the conference is adjusted to a media stream scattered distribution mode, including: the SIP application module sends a notification message to each terminal in the conference, each terminal in the conference returns an acknowledgement message after receiving the notification message, and modifies the sending address and port and the receiving address and port of the terminal media stream according to the parameter information of the notification message, changes the centralized media stream transmission mode to a distributed media stream transmission mode, and releases the edge MCU resource.

[0061] The present application effectively combines the centralized and distributed media stream transmission modes in the media stream control method, and effectively plays the respective advantages of the two control methods.

[0062] The intelligent control module collects the performance parameters of the edge media server in real time; the performance parameters of the edge media server include CPU usage, memory occupation, current call number, codec queue length, and bandwidth utilization; and a trained prediction model is used to predict the load prediction of the edge media server. Specifically, the method comprises:

[0063] A data set is obtained, and the data set includes the performance parameters of the edge media server. The data set is divided into a training set, a validation set, and a test set. The training set is used to train the model, and the performance of the model is evaluated on the validation set. If the loss function value on the validation set converges to a stable value, and continues to train without significant decline, the training is stopped.

[0064] Figure 6 A schematic diagram of the prediction model is shown in FIG. 1. Figure 6As shown, the prediction model comprises a preprocessing module, an LSTM layer, an attention mechanism layer, a fully connected layer, an activation function and an output layer. The preprocessing module aggregates, denoises and normalizes the training set data by time window to generate a feature matrix; the feature matrix is input into the LSTM layer to capture short-term local time series fluctuations; the attention mechanism layer is used to identify key time points (such as burst traffic peaks) so that the model pays more attention to these key information; the fully connected layer is used for global feature integration to output the load prediction for a future preset time length (such as 5s). The activation function is used to capture the complex interaction between the above performance parameters and enhance the non-linear expression ability. The output layer is used to directly output the load value measuring the load size.

[0065] Embodiment 2: The present application provides an intelligent and digital fusion communication system based on artificial intelligence, comprising a central service platform and a plurality of domains; wherein the central service platform comprises a location server and a central media server; the domains comprise a conference server, a WebRTC server, a gateway, a client and an edge media server; the client comprises a SIP client and a WebRTC client.

[0066] The conference server comprises a SIP application module and an intelligent and digital control module.

[0067] The gateway is located between the conference server and the WebRTC server, and is used to convert WebRTC signaling and SIP signaling to each other to complete mutual communication between heterogeneous networks.

[0068] The SIP terminal is directly connected to the conference server, and the WebRTC terminal is connected to the conference server via the WebRTC server and the gateway.

[0069] The location server is used to record and maintain the domain information of the global user.

[0070] The central media server is used to take over part of the media stream of the edge media server when the load of the edge media server in each domain is too large.

[0071] The present application has the following beneficial effects:

[0072] (1) The prediction model uses performance parameters such as CPU usage and memory occupancy of the edge media server to predict future load conditions. By knowing the load change of the edge media server in advance, the system can take measures before the load is too large. When the load of the edge media server is predicted to be too large, part of the media stream of the conference is transferred to the central media server or the media stream distribution mode is adjusted in advance to avoid performance degradation due to server overload, thereby realizing optimal allocation of resources.

[0073] (2) The use of a central media server can effectively alleviate the pressure on edge media servers, optimize resource allocation, and ensure smooth conference progress. This not only improves the overall performance of the system, but also meets the needs of conferences of different sizes. In actual conference scenarios, if multiple large conferences are conducted simultaneously, the edge media server may be overloaded due to processing too many media streams. By taking over the central media server, the edge media server can avoid problems such as freezing and crashing due to resource exhaustion, ensuring the stability of the entire system.

[0074] (3) By setting priorities for conferences, it helps to optimize resource allocation and ensure the quality and stability of important conferences. In centralized media stream transmission, the media stream processing and forwarding of high-priority conferences are prioritized to avoid problems such as audio and video freezing and delay due to insufficient resources, ensuring the smooth progress and high-quality communication of important conferences. For low-priority conferences with high resource occupancy density, by taking over some low-priority conference media streams by the central media server, it can reduce the load of the edge media server, provide high-quality services for high-priority conferences, and ensure the communication quality of conferences with high resource occupancy density. In small conferences with few people, distributed transmission can reduce intermediate links and transmission delay.

[0075] (4) This scheme combines distributed and centralized media stream distribution methods, which can fully utilize the advantages of both methods, improve the adaptability of the system in different scenarios, effectively alleviate server pressure, and ensure the quality of conference communication. According to the specific situation of the conference, the transmission method can be flexibly switched to ensure the quality of conference communication in different scenarios and avoid problems such as discontinuous images and audio flow interruption.

[0076] In summary, the AI-based digital fusion communication method and system provided in the present application combines SIP and WebRTC technologies to provide a comprehensive video conference solution, meeting the needs of users for high-quality, easy-to-use, and secure video communication.

[0077] The basic principles of the present disclosure are described above in conjunction with specific embodiments, but it should be noted that the advantages, advantages, and effects mentioned in the present disclosure are only examples and are not limiting, and these advantages, advantages, and effects cannot be considered as the must-have of each embodiment of the present disclosure. In addition, the above-mentioned specific details are only for the purpose of example and understanding, and are not limiting, and the above-mentioned details do not limit the present disclosure to the above-mentioned specific details.

[0078] Each of the embodiments described in this specification has at least one advantage in terms of the following aspects 1 to 10.

[0079] The block diagrams of the devices, apparatuses, equipment, systems involved in the present disclosure are only illustrative examples and are not intended to require or imply that the connection, arrangement, configuration must be as shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, equipment, systems can be connected, arranged, configured in any manner. Words such as "include", "contain", "have" and the like are open-ended words, mean "including but not limited to", and can be used interchangeably. The words "or" and "and" used herein mean the word "and / or", and can be used interchangeably unless the context clearly indicates otherwise. The word "such as" used herein means the phrase "such as but not limited to", and can be used interchangeably.

[0080] It should also be noted that in the devices, apparatuses and methods of the present disclosure, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions of the present disclosure. The above description of the disclosed aspects is provided so that any person skilled in the art can make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein, but is intended to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0081] The above description has been given for the purpose of illustration and description. Furthermore, this description does not intend to limit the embodiments of the present disclosure to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain modifications, alterations, changes, additions and sub-combinations thereof.

[0082] The above description is only the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A digital-intelligent converged communication method based on artificial intelligence, characterized in that: The SIP application module sets meeting priorities for each meeting hosted on the edge media server; the meeting clients include SIP terminals and / or WebRTC terminals; the SIP terminals are directly connected to the meeting server, and the WebRTC terminals are connected to the meeting server via a WebRTC server and a gateway; the meeting server includes the SIP application module and the digital control module. The intelligent control module sorts low-priority meetings from highest to lowest resource usage density, and obtains the top N low-priority meetings with the highest resource usage density; it also sorts low-priority meetings from lowest to highest number of participants, and obtains the top N low-priority meetings with the fewest participants; the value of N is preset. The intelligent control module collects performance parameters of the edge media server in real time; uses a trained prediction model to obtain the load prediction of the edge media server; and dynamically adjusts the media stream distribution mode based on the load prediction, specifically including: When the load prediction indicates excessive load, for the top N low-priority meetings with high resource consumption density, the central media server will take over the media stream of the meeting; for the top N low-priority meetings with few participants, the media stream distribution method of the meeting will be adjusted to a distributed media stream transmission method.

2. The method as described in claim 1, characterized in that: The prediction model includes: a preprocessing module, an LSTM layer, an attention mechanism layer, a fully connected layer, an activation function, and an output layer. The preprocessing module aggregates, denoises, and normalizes the training set data according to time windows to generate a feature matrix. The feature matrix is ​​input into the LSTM layer to capture short-term local temporal fluctuations. The attention mechanism layer is used to identify key time points. The fully connected layer is used to integrate global features and output a load prediction for a preset duration in the future. The activation function is used to capture the complex interaction relationships between the above performance parameters and enhance nonlinear expression capabilities. The output layer is used to directly output the load value that measures the load size.

3. The method as described in claim 1, characterized in that: The resource occupancy density is the number of participants divided by the bandwidth used by the meeting.

4. The method as described in claim 1, characterized in that: For the first N low-priority conferences with high resource consumption density, the central media server takes over the media stream of the conference. This includes: the SIP application module sending notification messages to the central media server and each terminal in the conference; the central media server and each terminal in the conference returning confirmation messages after receiving the notification messages; and each terminal in the conference modifying the sending address and port, as well as the receiving address and port of the terminal media stream, according to the parameter information in the notification message, so that the central media server takes over the media stream of the conference.

5. The method as described in claim 1, characterized in that: For the first N low-priority meetings with few participants, the media stream distribution method of the meeting is adjusted to a distributed media stream transmission method. This includes: the SIP application module sends a notification message to each terminal in the meeting; each terminal in the meeting returns an acknowledgment message after receiving the notification message; and modifies the sending address and port, as well as the receiving address and port of the terminal media stream, according to the parameter information of the notification message, thereby changing the centralized media stream transmission method into a distributed media stream transmission method.

6. The method as described in claim 1, characterized in that: The performance parameters of the edge media server include CPU utilization, memory usage, current number of calls, encoding / decoding queue length, and bandwidth utilization.

7. The method as described in claim 1, characterized in that: The SIP application module sets meeting priorities for each conference hosted by the edge media server, including: the SIP application module sets meeting priorities based on the client identifiers participating in the conference; when a specific client participates in the conference, the SIP application module automatically sets the conference to high priority.

8. A digital-intelligent converged communication system based on artificial intelligence, used to perform the method as described in any one of claims 1-7, characterized in that: include: A central service platform and multiple domains; wherein, the central service platform includes a location server and a central media server; the domains include a conferencing server, a WebRTC server, a gateway, clients, and edge media servers; the clients include SIP clients and WebRTC clients; The conference server includes a SIP application module and a digital control module; The gateway is located between the conference server and the WebRTC server and is used to convert WebRTC signaling to SIP signaling and to enable communication between heterogeneous networks. The SIP terminal connects directly to the conference server, while the WebRTC terminal connects to the conference server via a WebRTC server and a gateway. Location servers are used to record and maintain the domain information of all users globally; The central media server is used to take over a portion of the media streams from edge media servers in each domain when the load is too high.

Images