A semantic communication method, apparatus and system

Abstract

The application discloses a semantic communication method, device and system, wherein a user terminal is connected with a client front-end device and a semantic communication server; the client front-end device receives original data sent by the user terminal; the original data is subjected to semantic coding through a semantic coding model to obtain semantic coding data; and the semantic coding data is sent to the semantic communication server so that the semantic communication server subjects the semantic coding data to semantic decoding through a semantic decoding model corresponding to the semantic coding model. By deploying a semantic model on the client front-end device, the application uses the computing power and network resources of the client front-end device to realize semantic processing, reduces the deployment and maintenance cost of the semantic model of the user terminal, and improves user perception.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a semantic communication method, apparatus and system. Background Technology

[0002] For decades, research in the field of communications has primarily focused on how to accurately and efficiently transmit symbols from the sender to the receiver—a concept known as semantic communication. However, with the development of wireless communication systems and the accelerated implementation of intelligent and IoT applications, the capacity of wireless systems is gradually approaching the Shannon limit, posing an unprecedented challenge to traditional wireless communication capabilities. Therefore, semantic communication has emerged as one of the key potential technologies in communications. Unlike traditional communication, the main purpose of semantic communication is to achieve accurate interaction of semantic information; the transmitted portion consists of semantic information, which can significantly save bandwidth resources.

[0003] Semantic communication requires the use of semantic models to extract semantic information from transmitted data, which is then transmitted to the receiving end for decoding. Current technologies for semantic communication necessitate the deployment of semantic processing modules on user terminals. However, the sheer number of user terminals in the market (IoT devices, mobile phone users, etc.) makes it difficult to add semantic processing modules according to network operators' wishes, and the process is also characterized by high maintenance difficulty, high deployment costs, poor user experience, and limited adoption. Therefore, new approaches are needed to address these issues, reduce the bandwidth requirements of high-bandwidth services on private networks, and achieve both improved quality and reduced costs for private network services. Summary of the Invention

[0004] The purpose of this invention is to provide a semantic communication method, apparatus, and system that deploys a semantic model on a client-side device and utilizes the computing power and network resources of the client-side device to achieve semantic processing, thereby reducing the deployment and maintenance costs of the semantic model on the user terminal and improving user experience.

[0005] To achieve the above objectives, embodiments of the present invention provide a semantic communication method applied to a client front-end device, wherein a user terminal connects to a semantic communication server through the client front-end device; the semantic communication method includes: Receive raw data sent by the user terminal; The original data is semantically encoded using a semantic coding model to obtain semantically encoded data; The semantically encoded data is sent to the semantic communication server, so that the semantic communication server performs semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model.

[0006] As an improvement to the above scheme, the different semantic coding models of the client front-end device are used to process the semantic coding of data with different semantic communication needs, and each of the semantic coding models is predefined with its own first network access address. The step of semantically encoding the original data using a semantic coding model to obtain semantically encoded data includes: Based on the first network access address accessed by the user terminal, the original data is semantically encoded using the semantic coding model corresponding to the first network access address to obtain semantically encoded data; wherein, the user terminal can configure the first network access address to which the original data needs to be sent according to its own semantic communication requirements, and send the original data to the first network access address.

[0007] As an improvement to the above scheme, the semantic communication server has different semantic encoding models for processing semantic decoding of data with different semantic communication needs, and each semantic decoding model is predefined with its own second network access address. The step of sending the semantically encoded data to the semantic communication server, so that the semantic communication server performs semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model, includes: Based on the semantic encoding model accessed by the original data, a second network access address for the corresponding semantic decoding model is determined, and the semantic encoded data is sent to the second network access address so that the semantic communication server performs semantic decoding on the semantic encoded data through the semantic decoding model corresponding to the second network access address.

[0008] As an improvement to the above scheme, the first network access address and the second network access address are URL addresses.

[0009] As an improvement to the above scheme, the first network access address includes the IP address, port number, and corresponding semantic encoding model number of the client front-end device, and the second network access address includes the IP address, port number, and corresponding semantic decoding model number of the semantic communication server.

[0010] As an improvement to the above scheme, the semantic encoding model and the semantic decoding model are trained and shared by the semantic communication server.

[0011] As an improvement to the above solution, sending the semantically encoded data to the semantic communication server includes: The semantic encoded data is channel encoded to form channel encoded data and sent to the mobile communication base station, so that the mobile communication base station performs channel decoding on the channel encoded data to obtain the semantic encoded data and sends it to the semantic communication server.

[0012] This invention also provides another semantic communication method applied to a semantic communication server, wherein a user terminal connects to the semantic communication server via a client front-end device; the semantic communication method includes: Receive semantically encoded data sent by the client front-end device; wherein, the semantically encoded data is obtained by the client front-end device after semantically encoding the raw data sent by the user terminal through a semantic encoding model; The semantically encoded data is semantically decoded using the semantic decoding model corresponding to the semantic encoding model.

[0013] As an improvement to the above scheme, the semantic communication method further includes: Construct and train the semantic encoding model and the semantic decoding model; The semantic encoding model is shared to the client's front-end device for configuration.

[0014] As an improvement to the above scheme, the construction and training of the semantic encoding model and the semantic decoding model includes: A first neural network model for semantic encoding and a second neural network model for semantic decoding are constructed according to different semantic communication requirements, and the first neural network model is deployed to the client front-end device. Receive training data sent by the client front-end device; wherein the training data is sent from the user terminal to the client front-end device; The system receives semantic encoding training data sent by the client front-end device; wherein the semantic encoding training data is obtained by the client front-end device semantically encoding the training data through the first neural network model. The semantic encoding training data is semantically decoded using the second neural network model to obtain semantic decoding training data; The training data and the semantic decoding training data are compared, and the parameters of the first neural network model and the second neural network model are tuned according to the comparison results. Training ends when the accuracy of the first neural network model and the second neural network model reaches a preset accuracy threshold, or the number of training iterations reaches a preset number threshold. The semantic encoding model is obtained based on the first neural network model, and the semantic decoding model is obtained based on the second neural network model.

[0015] This invention also provides a semantic communication device applied to a client front-end device, through which a user terminal connects to a semantic communication server; the semantic communication device includes: The raw data receiving module is used to receive raw data sent by the user terminal; The semantic encoding module is used to perform semantic encoding on the original data through a semantic encoding model to obtain semantically encoded data. The data sending module is used to send the semantically encoded data to the semantic communication server, so that the semantic communication server can perform semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model.

[0016] This invention also provides another semantic communication device applied to a semantic communication server, wherein a user terminal connects to the semantic communication server via a client front-end device; the semantic communication device includes: The encoded data receiving module is used to receive semantically encoded data sent by the client front-end device; wherein, the semantically encoded data is obtained by the client front-end device after semantically encoding the raw data sent by the user terminal through a semantic encoding model; The semantic decoding module is used to perform semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model.

[0017] This invention also provides a semantic communication system, which includes a user terminal, a client front-end device, and a semantic communication server, wherein the user terminal is connected to the semantic communication server through the client front-end device. The user terminal is used to send raw data to the customer front-end device; The client front-end device is used to semantically encode the raw data through a semantic coding model to obtain semantically encoded data, and then send the semantically encoded data to the semantic communication server. The semantic communication server is used to perform semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model.

[0018] Compared with existing technologies, the semantic communication method, apparatus, and system disclosed in this invention can fully utilize the computing resources of communication computing networks. Through semantic communication, it significantly reduces the bandwidth requirements of private network users in customer front-end device scenarios, reduces the allocation of dedicated resources for private networks, and thus greatly reduces private network costs. Simultaneously, this invention overcomes the shortcomings of existing semantic communication methods in edge computing networks. By deploying semantic models on customer front-end devices to achieve semantic processing, it can rationally utilize the computing and network resources of customer front-end devices, improve the efficiency of computing network resource utilization, and eliminate the need for user terminals to install semantic processing modules. This improves the applicability of semantic communication, reduces the deployment and maintenance costs of semantic models on user terminals, and enhances user experience. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a semantic communication method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the semantic communication architecture under the edge computing network in an embodiment of the present invention; Figure 3 This is a schematic diagram of a partial semantic communication architecture based on URLs in an embodiment of the present invention; Figure 4 This is a flowchart illustrating another semantic communication method provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a semantic communication device provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of another semantic communication device provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of a semantic communication system provided in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See Figure 1 This is a flowchart illustrating a semantic communication method provided in an embodiment of the present invention. The embodiment provides a semantic communication method applied to a client front-end device, through which a user terminal connects to a semantic communication server. The semantic communication method includes steps S11 to S13: S11. Receive the raw data sent by the user terminal; S12. The original data is semantically encoded using a semantic coding model to obtain semantically encoded data; S13. The semantic encoded data is sent to the semantic communication server so that the semantic communication server performs semantic decoding on the semantic encoded data through the semantic decoding model corresponding to the semantic encoding model.

[0022] It should be noted that the embodiments of the present invention are applied to mobile communication edge computing networks. In edge computing networks, customer-premise equipment (CPEs) are deployed extensively on the user access side in various scenarios such as hospitals, institutions, and factories. These CPEs provide user access while also offering powerful edge computing capabilities at the closest possible distance. CPE technology is constantly evolving, and the market size is continuously expanding. Currently, mainstream products support new technologies such as Wi-Fi 6, enabling them to provide higher transmission rates. They are also equipped with high-performance processors and large memory capacities, making them important computing nodes in edge computing networks and demonstrating advanced technology.

[0023] Therefore, in this embodiment of the invention, the semantic processing module is deployed on the customer's front-end device, and the computing power and network resources of the customer's front-end device are used to realize semantic processing, providing semantic processing capabilities for private network terminals and reducing the user terminal's demand for mobile communication network bandwidth.

[0024] See Figure 2 This is a schematic diagram of the semantic communication architecture under the edge computing network in this embodiment of the invention. In the edge computing network, the customer front-end device 30 is connected to the user terminal system wireless local area network downwards and to the mobile communication base station 20 and semantic communication server 10 upwards. The customer front-end device 30 plays a connecting role.

[0025] In this embodiment of the invention, the semantic communication server 10 is deployed behind the mobile communication base station 20. The specific location depends on the actual situation; it can be deployed at a node in the transmission network or in the core network. Its relationship with the mobile communication base station 20 is one-to-many. The semantic communication server 10 is equipped with at least an encoded data receiving module 11 and a semantic decoding module 12. The client front-end device 30 provides edge computing power and network connectivity, connects to the semantic communication server 10 via a tunneling protocol, and is equipped with at least a raw data receiving module 31, a semantic encoding module 32, and a data sending module 33. The user terminal 40 is a semantic communication terminal in the edge computing network, used to send raw data.

[0026] The semantic encoding module 32 is configured with different types of semantic encoding models to implement semantic encoding of data with different semantic communication needs. The semantic decoding module 12 is configured with different types of semantic decoding models to implement semantic decoding of data with different semantic communication needs and corresponding semantic encoding methods. The semantic encoding model and the semantic decoding model cooperate with each other to perform model training.

[0027] In this embodiment of the invention, the raw data received by the raw data receiving module 31 of the client front-end device 30 from the user terminal 40 is input into the corresponding semantic encoding module 32. The semantic encoding module 32 uses a configured semantic encoding model to semantically encode the raw data to obtain semantically encoded data, and sends the semantically encoded data to the semantic communication server 10 through the data sending module 33. The semantic communication server 10 receives the semantically encoded data through the encoded data receiving module 11, and uses the semantic encoding module 12 to semantically decode the semantically encoded data using a configured semantic decoding model corresponding to the semantic encoding model and outputs the result.

[0028] Optionally, sending the semantically encoded data to the semantic communication server specifically involves: The semantic encoded data is channel encoded to form channel encoded data, and then sent to the mobile communication base station, so that the mobile communication base station can perform channel decoding on the channel encoded data to obtain the semantic encoded data, and then send it to the semantic communication server.

[0029] In this embodiment of the invention, the mobile communication base station 20 is at least used to forward semantically encoded data sent by the client front-end device 30 to the semantic communication server 10. See also Figure 2 The mobile communication air interface refers to the air interface between the mobile communication base station 20 and the customer front-end equipment 30. The purpose of edge computing network semantic communication is to reduce the demand of user terminals for this wireless resource. The wireless LAN air interface refers to the air interface between the customer front-end equipment 30 and the user terminal 40.

[0030] After the client front-end device 30 processes the semantically encoded data, it uses its data transmission module 33, such as a 5G mobile communication module, to perform traditional channel coding and send the data to the mobile communication base station 20. The mobile communication base station receives the data and performs channel decoding, and then transmits it to the semantic communication server 10.

[0031] By employing the technical means of this invention, the present invention can fully utilize the computing resources of the communication computing network, and greatly reduce the bandwidth requirements of private network users in customer front-end device scenarios through semantic communication, thereby reducing the allocation of dedicated resources for the private network and significantly lowering the cost of the private network. Simultaneously, this invention overcomes the shortcomings of existing semantic communication methods in edge computing networks. By deploying a semantic model on the customer front-end device to achieve semantic processing, it can rationally utilize the computing and network resources of the customer front-end device, improve the efficiency of computing network resource utilization, and eliminate the need for user terminals to install semantic processing modules. This improves the applicability of semantic communication, reduces the deployment and maintenance costs of semantic models for user terminals, and enhances user experience.

[0032] As a preferred embodiment, the present invention is further implemented based on the above embodiments, wherein the semantic encoding model and the semantic decoding model are trained and shared by the semantic communication server 10.

[0033] Specifically, the semantic communication server 10 also deploys a semantic sharing module 13. The semantic sharing module 13 is used to construct a semantic knowledge base, process semantic content including but not limited to video, audio, images, and text, train semantic encoding and decoding models for the aforementioned content, deploy the semantic encoding model in the semantic encoding module 32 of the client front-end device 30, and deploy the semantic decoding model in the semantic decoding module 12. Simultaneously, based on the federated learning framework, different semantic communication servers can directly share model parameters between encoding and decoding models performing the same semantic processing task. That is, the semantic sharing module 13 can also share the trained model parameters to other semantic communication servers 10 with the same semantic processing task. After sharing, each of the other semantic communication servers optimizes its semantic encoding and decoding model based on its own semantic knowledge background.

[0034] Preferably, the semantic communication server 10 trains the semantic encoding model and the semantic decoding model through the following steps: A first neural network model for semantic encoding and a second neural network model for semantic decoding are constructed according to different semantic communication requirements, and the first neural network model is deployed to the client front-end device. Receive training data sent by the client front-end device; wherein the training data is sent from the user terminal to the client front-end device; The system receives semantic encoding training data sent by the client front-end device; wherein the semantic encoding training data is obtained by the client front-end device semantically encoding the training data through the first neural network model. The semantic encoding training data is semantically decoded using the second neural network model to obtain semantic decoding training data; The training data and the semantic decoding training data are compared, and the parameters of the first neural network model and the second neural network model are tuned according to the comparison results. Training ends when the accuracy of the first neural network model and the second neural network model reaches a preset accuracy threshold, or the number of training iterations reaches a preset number threshold. The semantic encoding model is obtained based on the first neural network model, and the semantic decoding model is obtained based on the second neural network model.

[0035] In this embodiment of the invention, the semantic sharing module 13 of the semantic communication server 10 trains the semantic encoding model and the semantic decoding model based on its semantic sharing knowledge background, and shares the model parameters with other semantic sharing modules under the semantic communication server that perform the same semantic communication processing task but have different knowledge backgrounds through a federated learning framework. The shared semantic sharing modules deploy the semantic encoding model to the semantic decoding module 32 of the client front-end device, and the semantic decoding model to the semantic decoding module 12 of the semantic communication server.

[0036] During model training, the user terminal 40 corresponding to the semantic encoding / decoding model sends the original training data to the client front-end device 30. The semantic encoding model is used to encode and extract semantic information, which is then transmitted via the mobile communication air interface to the semantic decoding module 12 of the semantic communication server 10 through channel encoding / decoding for semantic decoding. The decoding result is then sent to the semantic sharing module 13. Additionally, the client front-end device 40 also directly sends the original training data to the semantic sharing module 13 of the semantic communication server 10.

[0037] The semantic sharing module 13 compares the original training data with the semantic decoding training data to determine the correctness of the output results of the semantic encoding and decoding models, and optimizes the parameters of the two models based on the results. Training ends when the model accuracy reaches a threshold or the number of iterations exceeds a threshold, thus completing the training of the semantic encoding model and semantic decoding model of the semantic communication server.

[0038] Optionally, the semantic encoding model and the semantic decoding model may include, but are not limited to, CNN, seq2seq, transformer, BERT, SERT, etc.

[0039] By employing the technical means of this invention, a semantic sharing module is set up on the semantic communication server to realize the training, deployment, and sharing of semantic encoding and decoding models. This enables the encoding and decoding of data of various different business types and improves the efficiency of computing network resource utilization.

[0040] As a preferred embodiment, the present invention is further implemented based on the above embodiments. The different semantic coding models of the client front-end device 30 are used to process the semantic coding of data with different semantic communication needs, and each of the semantic coding models is predefined with its own first network access address. The semantic communication server 10 has different semantic encoding models for processing semantic decoding of data with different semantic communication needs, and each semantic decoding model has its own predefined second network access address.

[0041] In step S12, the step of semantically encoding the original data using a semantic coding model to obtain semantically encoded data includes: Based on the first network access address accessed by the user terminal, the original data is semantically encoded using the semantic coding model corresponding to the first network access address to obtain semantically encoded data; wherein, the user terminal can configure the first network access address to which the original data needs to be sent according to its own semantic communication requirements, and send the original data to the first network access address.

[0042] In step S13, sending the semantically encoded data to the semantic communication server so that the semantic communication server performs semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model includes: Based on the semantic encoding model accessed by the original data, a second network access address for the corresponding semantic decoding model is determined, and the semantic encoded data is sent to the second network access address so that the semantic communication server performs semantic decoding on the semantic encoded data through the semantic decoding model corresponding to the second network access address.

[0043] In this embodiment of the invention, the semantic encoding module 32 deploys multiple semantic encoding models for handling different semantic communication needs, and the semantic decoding module 12 deploys multiple semantic decoding models for handling different semantic communication needs. Therefore, by defining separate network access addresses for different semantic encoding and decoding models, the encoding and decoding of raw data for different semantic communication needs can be accurately achieved.

[0044] Preferably, the first network access address and the second network access address are URL (Uniform Resource Locator) addresses.

[0045] Preferably, the first network access address includes the IP address, port number, and corresponding semantic encoding model number of the client front-end device, and the second network access address includes the IP address, port number, and corresponding semantic decoding model number of the semantic communication server.

[0046] See Figure 3 This is a schematic diagram of a partial semantic communication architecture based on URLs in an embodiment of the present invention. The semantic sharing module 13 deployed on the semantic communication server 10 will configure various semantic encoding and decoding models according to the semantic communication requirements of the base station it serves. The network access address of the semantic decoding model is defined on the semantic communication server 10 as the IP address of the semantic communication server 10 plus the port number, plus the English number of the semantic decoding model, forming a unique network access address. Similarly, the network access address of the raw data receiving module 31 and the semantic encoding model is defined on the client front-end device 30 as the IP address of the client front-end device 30 plus the port number, plus the English number of the semantic encoding model, forming a unique access address.

[0047] As an example, if the user terminal 40 is a surveillance camera used to detect intrusion, then the URL address for sending its raw data, which is also the first network access address, is configured as http: / / 192.168.10.1:5250 / video_intrusion_detection. Here, 192.168.10.1 is the IP address of the client front-end device 30, which is also the gateway address of the wireless local area network of the user terminal 40, 5250 is the application port number, and video_intrusion_detection is the English number of the semantic coding model.

[0048] The semantic encoding model in the semantic encoding module 32 deployed on the client front-end device 30 is connected to the corresponding semantic decoding model under the semantic decoding module 12 of the semantic communication server 10. That is, the semantic encoding module is configured with the URL address of the corresponding semantic decoding model, which is also the second network access address, http: / / 10.227.4.48:5250 / video_intrusion_detection, where 10.227.4.48 is the IP address of the semantic communication server 10, 5250 is the application port number, and video_intrusion_detection is the English number of the semantic decoding model.

[0049] In practical applications, the user terminal 40 configures the URL address for sending the raw data, i.e., the first network access address, according to its semantic communication requirements, and sends the raw data to that URL address. The raw data receiving module 31 in the client front-end device 30 receives the raw data, processes it in real time according to the requirements of the semantic coding model, and inputs the raw data into the semantic coding module 32 for semantic coding. The semantic coding module 32 uses a pre-trained semantic coding model to perform semantic coding on the raw data and determines the corresponding URL address, i.e., the second network access address. After completing the coding, the semantic coding module 32 immediately sends the semantically encoded data to the semantic decoding module 12 on the semantic communication server 10. The semantic decoding module 12 uses a pre-trained semantic decoding model corresponding to the semantic coding model to perform semantic decoding and output the result.

[0050] This invention employs the technical means of embodiments of the present invention to establish a URL-based semantic communication network architecture. On the client front-end device and semantic communication server sides, semantic encoding and decoding models are paired via URLs, enabling the deployment of multiple semantic processing models to meet the encoding and decoding needs of data with different semantic communication requirements. User terminals execute semantic communication tasks by configuring URLs, achieving adaptive semantic communication. This invention can fully utilize the computing resources of the computing network, significantly reducing the bandwidth requirements of private network users in client front-end device scenarios through semantic communication, reducing the allocation of dedicated resources for private networks, thereby greatly reducing private network costs and facilitating industry support for market promotion.

[0051] See Figure 4 This is a flowchart illustrating another semantic communication method provided in an embodiment of the present invention. The present invention also provides a semantic communication method applied to a semantic communication server, wherein a user terminal connects to the semantic communication server through a client front-end device.

[0052] The semantic communication method includes steps S21 to S22: S21. Receive semantically encoded data sent by the client front-end device; wherein, the semantically encoded data is obtained by the client front-end device after semantically encoding the original data sent by the user terminal through a semantic encoding model; S22. The semantically encoded data is semantically decoded using the semantic decoding model corresponding to the semantic encoding model.

[0053] It should be noted that this embodiment of the invention is applied to a mobile communication edge computing network. In the edge computing network, the client front-end device 30 is connected downwards to the user terminal system's wireless local area network and upwards to the mobile communication base station 20 and the semantic communication server 10. The semantic communication server 10 is equipped with at least an encoded data receiving module 11 and a semantic decoding module 12. The client front-end device 30 is used to provide edge computing power and network, connects to the semantic communication server 10 through a tunneling protocol, and is equipped with at least a raw data receiving module 31, a semantic encoding module 32, and a data sending module 33. The user terminal 40 is a semantic communication terminal in the edge computing network, used to send raw data.

[0054] The semantic encoding module 32 is configured with different types of semantic encoding models to implement semantic encoding of data with different semantic communication needs. The semantic decoding module 12 is configured with different types of semantic decoding models to implement semantic decoding of data with different semantic communication needs and corresponding semantic encoding methods. The semantic encoding model and the semantic decoding model cooperate with each other to perform model training.

[0055] In this embodiment of the invention, the raw data received by the raw data receiving module 31 of the client front-end device 30 from the user terminal 40 is input into the corresponding semantic encoding module 32. The semantic encoding module 32 uses a configured semantic encoding model to semantically encode the raw data to obtain semantically encoded data, and sends the semantically encoded data to the semantic communication server 10 through the data sending module 33. The semantic communication server 10 receives the semantically encoded data through the encoded data receiving module 11, and uses the semantic encoding module 12 to semantically decode the semantically encoded data using a configured semantic decoding model corresponding to the semantic encoding model and outputs the result.

[0056] Preferably, the semantic communication server 10 is connected to the client front-end device 30 via the mobile communication base station 20. Then, receiving the semantically encoded data sent by the client front-end device includes: The system receives semantically encoded data sent by the mobile communication base station after channel decoding of the channel-coded data; wherein the channel-coded data is formed by the client front-end device performing channel coding on the semantically encoded data, and the semantically encoded data is obtained by the client front-end device performing semantic coding on the original data sent by the user terminal through a semantic coding model.

[0057] Specifically, after the client front-end device 30 processes and obtains the semantic encoded data, it uses its data transmission module 33, such as a 5G mobile communication module, to perform traditional channel coding and send the data to the mobile communication base station 20. The mobile communication base station receives the data and performs channel decoding, and then transmits it to the semantic communication server 10.

[0058] By employing the technical means of this invention, the present invention can fully utilize the computing resources of the communication computing network, and greatly reduce the bandwidth requirements of private network users in customer front-end device scenarios through semantic communication, thereby reducing the allocation of dedicated resources for the private network and significantly lowering the cost of the private network. Simultaneously, this invention overcomes the shortcomings of existing semantic communication methods in edge computing networks. By deploying a semantic model on the customer front-end device to achieve semantic processing, it can rationally utilize the computing and network resources of the customer front-end device, improve the efficiency of computing network resource utilization, and eliminate the need for user terminals to install semantic processing modules. This improves the applicability of semantic communication, reduces the deployment and maintenance costs of semantic models for user terminals, and enhances user experience.

[0059] As a preferred embodiment, the present invention further implements the above embodiments, and the semantic communication method further includes the following steps: Construct and train the semantic encoding model and the semantic decoding model; share the semantic encoding model to the client front-end device for configuration.

[0060] In this embodiment of the invention, the semantic communication server 10 is further deployed with a semantic sharing module 13. The semantic sharing module 13 is used to construct a semantic knowledge base, process semantic content including but not limited to video, audio, images, and text, train semantic encoding and decoding models for the aforementioned content, deploy the semantic encoding model in the semantic encoding module 32 of the client front-end device 30, and deploy the semantic decoding model in the semantic decoding module 12. Simultaneously, based on the federated learning framework, different semantic communication servers can directly share model parameters between encoding and decoding models performing the same semantic processing task. That is, the semantic sharing module 13 can also share the trained model parameters to other semantic communication servers 10 with the same semantic processing task. After sharing, each of the other semantic communication servers optimizes its semantic encoding and decoding model based on its own semantic knowledge background.

[0061] Preferably, constructing and training the semantic encoding model and the semantic decoding model includes: A first neural network model for semantic encoding and a second neural network model for semantic decoding are constructed according to different semantic communication requirements, and the first neural network model is deployed to the client front-end device. Receive training data sent by the client front-end device; wherein the training data is sent from the user terminal to the client front-end device; The system receives semantic encoding training data sent by the client front-end device; wherein the semantic encoding training data is obtained by the client front-end device semantically encoding the training data through the first neural network model. The semantic encoding training data is semantically decoded using the second neural network model to obtain semantic decoding training data; The training data and the semantic decoding training data are compared, and the parameters of the first neural network model and the second neural network model are tuned according to the comparison results. Training ends when the accuracy of the first neural network model and the second neural network model reaches a preset accuracy threshold, or the number of training iterations reaches a preset number threshold. The semantic encoding model is obtained based on the first neural network model, and the semantic decoding model is obtained based on the second neural network model.

[0062] In this embodiment of the invention, the semantic sharing module 13 of the semantic communication server 10 trains the semantic encoding model and the semantic decoding model based on its semantic sharing knowledge background, and shares the model parameters with other semantic sharing modules under the semantic communication server that perform the same semantic communication processing task but have different knowledge backgrounds through a federated learning framework. The shared semantic sharing modules deploy the semantic encoding model to the semantic decoding module 32 of the client front-end device, and the semantic decoding model to the semantic decoding module 12 of the semantic communication server.

[0063] During model training, the user terminal 40 corresponding to the semantic encoding / decoding model sends the original training data to the client front-end device 30. The semantic encoding model is used to encode and extract semantic information, which is then transmitted via the mobile communication air interface to the semantic decoding module 12 of the semantic communication server 10 through channel encoding / decoding for semantic decoding. The decoding result is then sent to the semantic sharing module 13. Additionally, the client front-end device 40 also directly sends the original training data to the semantic sharing module 13 of the semantic communication server 10.

[0064] The semantic sharing module 13 compares the original training data with the semantic decoding training data to determine the correctness of the output results of the semantic encoding and decoding models, and optimizes the parameters of the two models based on the results. Training ends when the model accuracy reaches a threshold or the number of iterations exceeds a threshold, thus completing the training of the semantic encoding model and semantic decoding model of the semantic communication server.

[0065] Optionally, the semantic encoding model and the semantic decoding model may include, but are not limited to, CNN, seq2seq, transformer, BERT, SERT, etc.

[0066] By employing the technical means of this invention, a semantic sharing module is set up on the semantic communication server to realize the training, deployment, and sharing of semantic encoding and decoding models. This enables the encoding and decoding of data of various different business types and improves the efficiency of computing network resource utilization.

[0067] As a preferred embodiment, the present invention further implements the above embodiments, wherein different semantic encoding models of the client front-end device are used to process the semantic encoding of data with different semantic communication needs, and each semantic encoding model is predefined with its own first network access address; different semantic encoding models of the semantic communication server are used to process the semantic decoding of data with different semantic communication needs, and each semantic decoding model is predefined with its own second network access address.

[0068] During semantic communication, the user terminal configures a first network access address to which the original data needs to be sent according to its own semantic communication requirements, and sends the original data to the first network access address. The client front-end device, based on the first network access address accessed by the user terminal, performs semantic encoding on the original data using the semantic encoding model corresponding to the first network access address, obtaining semantically encoded data. The client front-end device then determines a second network access address for the corresponding semantic decoding model based on the semantic encoding model accessed by the original data, and sends the semantically encoded data to the second network access address. The semantic communication server performs semantic decoding on the semantically encoded data using the semantic decoding model corresponding to the second network access address.

[0069] Preferably, the first network access address and the second network access address are URL addresses.

[0070] Preferably, the first network access address includes the IP address, port number, and corresponding semantic encoding model number of the client front-end device, and the second network access address includes the IP address, port number, and corresponding semantic decoding model number of the semantic communication server.

[0071] This invention employs the technical means of embodiments of the present invention to establish a URL-based semantic communication network architecture. On the client front-end device and semantic communication server sides, semantic encoding and decoding models are paired via URLs, enabling the deployment of multiple semantic processing models to meet the encoding and decoding needs of data with different semantic communication requirements. User terminals execute semantic communication tasks by configuring URLs, achieving adaptive semantic communication. This invention can fully utilize the computing resources of the computing network, significantly reducing the bandwidth requirements of private network users in client front-end device scenarios through semantic communication, reducing the allocation of dedicated resources for private networks, thereby greatly reducing private network costs and facilitating industry support for market promotion.

[0072] See Figure 5 This is a schematic diagram of a semantic communication device provided in an embodiment of the present invention. The semantic communication device provided in this embodiment is applied to a client front-end device 30, and a user terminal connects to a semantic communication server through the client front-end device. The semantic communication device includes: The raw data receiving module 31 is used to receive raw data sent by the user terminal; Semantic encoding module 32 is used to semantically encode the original data through a semantic encoding model to obtain semantically encoded data; The data sending module 33 is used to send the semantic encoded data to the semantic communication server, so that the semantic communication server can perform semantic decoding on the semantic encoded data through the semantic decoding model corresponding to the semantic encoding model.

[0073] By employing the technical means of this invention, the present invention can fully utilize the computing resources of the communication computing network, and greatly reduce the bandwidth requirements of private network users in customer front-end device scenarios through semantic communication, thereby reducing the allocation of dedicated resources for the private network and significantly lowering the cost of the private network. Simultaneously, this invention overcomes the shortcomings of existing semantic communication methods in edge computing networks. By deploying a semantic model on the customer front-end device to achieve semantic processing, it can rationally utilize the computing and network resources of the customer front-end device, improve the efficiency of computing network resource utilization, and eliminate the need for user terminals to install semantic processing modules. This improves the applicability of semantic communication, reduces the deployment and maintenance costs of semantic models for user terminals, and enhances user experience.

[0074] It should be noted that the semantic communication device provided in this embodiment of the invention is used to execute all the process steps of a semantic communication method applied to a client front-end device in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0075] See Figure 6This is a schematic diagram of another semantic communication device provided in an embodiment of the present invention. This semantic communication device is applied to a semantic communication server 10, and a user terminal connects to the semantic communication server through the client front-end device. The semantic communication device includes: The encoded data receiving module 11 is used to receive semantically encoded data sent by the client front-end device; wherein, the semantically encoded data is obtained by the client front-end device after semantically encoding the original data sent by the user terminal through a semantic encoding model; The semantic decoding module 12 is used to perform semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model.

[0076] The device further includes: The semantic sharing module 13 is used to construct and train the semantic encoding model and the semantic decoding model; and to share the semantic encoding model to the client front-end device for configuration.

[0077] It should be noted that the semantic communication device provided in this embodiment of the invention is used to execute all the process steps of a semantic communication method applied to a semantic communication server in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0078] See Figure 7 This is a schematic diagram of the structure of a semantic communication system provided in an embodiment of the present invention. The embodiment of the present invention also provides a semantic communication system, which includes a user terminal 40, a client front-end device 30 and a semantic communication server 10. The user terminal 40 is connected to the semantic communication server 10 through the client front-end device 30.

[0079] The user terminal 40 is used to send raw data to the customer front-end device; The client front-end device 30 is used to perform semantic encoding on the original data through a semantic encoding model to obtain semantic encoded data, and send the semantic encoded data to the semantic communication server; The semantic communication server 10 is used to perform semantic decoding on the semantic encoded data through the semantic decoding model corresponding to the semantic encoding model.

[0080] Preferably, the semantic communication system further includes a mobile communication base station 20, and the client front-end device 30 is connected to the semantic communication server 10 through the mobile communication base station 20.

[0081] The client front-end device 30 performs channel coding on the semantic coding data to form channel-coded data, and sends it to the mobile communication base station 20. The mobile communication base station 20 performs channel decoding on the channel-coded data to obtain the semantic coding data, and sends it to the semantic communication server 10.

[0082] This invention is applied to a mobile communication edge computing network. In this edge computing network, the client front-end device 30 is connected downwards to the user terminal system's wireless local area network and upwards to the mobile communication base station 20 and the semantic communication server 10. The semantic communication server 10 is equipped with a semantic decoding module 12. The client front-end device 30 provides edge computing power and network connectivity, connects to the semantic communication server 10 via a tunneling protocol, and is equipped with at least a raw data receiving module 31, a semantic encoding module 32, and a data sending module 33. The user terminal 40 is a semantic communication terminal in the edge computing network, used to send raw data.

[0083] The semantic encoding module 32 is configured with different types of semantic encoding models to implement semantic encoding of data with different semantic communication needs. The semantic decoding module 12 is configured with different types of semantic decoding models to implement semantic decoding of data with different semantic communication needs and corresponding semantic encoding methods. The semantic encoding model and the semantic decoding model cooperate with each other to perform model training.

[0084] In this embodiment of the invention, the raw data received by the raw data receiving module 31 of the client front-end device 30 from the user terminal 40 is input into the corresponding semantic encoding module 32. The semantic encoding module 32 uses a configured semantic encoding model to semantically encode the raw data to obtain semantically encoded data, and sends the semantically encoded data to the semantic communication server 10 through the data sending module 33. The semantic communication server 10 receives the semantically encoded data through the encoded data receiving module 11, and uses the semantic encoding module 12 to semantically decode the semantically encoded data using a configured semantic decoding model corresponding to the semantic encoding model and outputs the result.

[0085] Preferably, the client front-end device 30 is connected to the semantic communication server 10 via the mobile communication base station 20. Then, the client front-end device sends the semantically encoded data to the semantic communication server, including: The client front-end device performs channel coding on the semantically encoded data to form channel-coded data, and sends it to the mobile communication base station; The mobile communication base station performs channel decoding on the channel-coded data to obtain the semantic-coded data, and then sends it to the semantic communication server.

[0086] Specifically, after the client front-end device 30 processes and obtains the semantic encoded data, it uses its data transmission module 33, such as a 5G mobile communication module, to perform traditional channel coding and send the data to the mobile communication base station 20. The mobile communication base station receives the data and performs channel decoding, and then transmits it to the semantic communication server 10.

[0087] By employing the technical means of this invention, the present invention can fully utilize the computing resources of the communication computing network, and greatly reduce the bandwidth requirements of private network users in customer front-end device scenarios through semantic communication, thereby reducing the allocation of dedicated resources for the private network and significantly lowering the cost of the private network. Simultaneously, this invention overcomes the shortcomings of existing semantic communication methods in edge computing networks. By deploying a semantic model on the customer front-end device to achieve semantic processing, it can rationally utilize the computing and network resources of the customer front-end device, improve the efficiency of computing network resource utilization, and eliminate the need for user terminals to install semantic processing modules. This improves the applicability of semantic communication, reduces the deployment and maintenance costs of semantic models for user terminals, and enhances user experience.

[0088] As a preferred embodiment, the present invention further implements the above embodiments, wherein the semantic communication server 10 is further configured to: construct and train the semantic encoding model and the semantic decoding model; and share the semantic encoding model to the client front-end device for configuration.

[0089] Specifically, the semantic communication server 10 also deploys a semantic sharing module 13. The semantic sharing module 13 is used to construct a semantic knowledge base, process semantic content including but not limited to video, audio, images, and text, train semantic encoding and decoding models for the aforementioned content, deploy the semantic encoding model in the semantic encoding module 32 of the client front-end device 30, and deploy the semantic decoding model in the semantic decoding module 12. Simultaneously, based on the federated learning framework, different semantic communication servers can directly share model parameters between encoding and decoding models performing the same semantic processing task. That is, the semantic sharing module 13 can also share the trained model parameters to other semantic communication servers 10 with the same semantic processing task. After sharing, each of the other semantic communication servers optimizes its semantic encoding and decoding model based on its own semantic knowledge background.

[0090] Preferably, constructing and training the semantic encoding model and the semantic decoding model includes: A first neural network model for semantic encoding and a second neural network model for semantic decoding are constructed according to different semantic communication requirements, and the first neural network model is deployed to the client front-end device. Receive training data sent by the client front-end device; wherein the training data is sent from the user terminal to the client front-end device; The system receives semantic encoding training data sent by the client front-end device; wherein the semantic encoding training data is obtained by the client front-end device semantically encoding the training data through the first neural network model. The semantic encoding training data is semantically decoded using the second neural network model to obtain semantic decoding training data; The training data and the semantic decoding training data are compared, and the parameters of the first neural network model and the second neural network model are tuned according to the comparison results. Training ends when the accuracy of the first neural network model and the second neural network model reaches a preset accuracy threshold, or the number of training iterations reaches a preset number threshold. The semantic encoding model is obtained based on the first neural network model, and the semantic decoding model is obtained based on the second neural network model.

[0091] In this embodiment of the invention, the semantic sharing module 13 of the semantic communication server 10 trains the semantic encoding model and the semantic decoding model based on its semantic sharing knowledge background, and shares the model parameters with other semantic sharing modules under the semantic communication server that perform the same semantic communication processing task but have different knowledge backgrounds through a federated learning framework. The shared semantic sharing modules deploy the semantic encoding model to the semantic decoding module 32 of the client front-end device, and the semantic decoding model to the semantic decoding module 12 of the semantic communication server.

[0092] During model training, the user terminal 40 corresponding to the semantic encoding / decoding model sends the original training data to the client front-end device 30. The semantic encoding model is used to encode and extract semantic information, which is then transmitted via the mobile communication air interface to the semantic decoding module 12 of the semantic communication server 10 through channel encoding / decoding for semantic decoding. The decoding result is then sent to the semantic sharing module 13. Additionally, the client front-end device 40 also directly sends the original training data to the semantic sharing module 13 of the semantic communication server 10.

[0093] The semantic sharing module 13 compares the original training data with the semantic decoding training data to determine the correctness of the output results of the semantic encoding and decoding models, and optimizes the parameters of the two models based on the results. Training ends when the model accuracy reaches a threshold or the number of iterations exceeds a threshold, thus completing the training of the semantic encoding model and semantic decoding model of the semantic communication server.

[0094] Optionally, the semantic encoding model and the semantic decoding model may include, but are not limited to, CNN, seq2seq, transformer, BERT, SERT, etc.

[0095] By employing the technical means of this invention, a semantic sharing module is set up on the semantic communication server to realize the training, deployment, and sharing of semantic encoding and decoding models. This enables the encoding and decoding of data of various different business types and improves the efficiency of computing network resource utilization.

[0096] As a preferred embodiment, the present invention further implements the above embodiments, wherein different semantic encoding models of the client front-end device are used to process the semantic encoding of data with different semantic communication needs, and each semantic encoding model is predefined with its own first network access address; different semantic encoding models of the semantic communication server are used to process the semantic decoding of data with different semantic communication needs, and each semantic decoding model is predefined with its own second network access address.

[0097] The user terminal sends the raw data to the client front-end device, including: the user terminal configuring the first network access address to which the raw data needs to be sent according to its own semantic communication requirements, and sending the raw data to the first network access address.

[0098] The client front-end device performs semantic encoding on the original data through a semantic encoding model to obtain semantically encoded data, including: the client front-end device performs semantic encoding on the original data through a semantic encoding model corresponding to the first network access address accessed by the user terminal to obtain semantically encoded data; The client front-end device sends the semantic encoded data to the semantic communication server, including: the client front-end device determines the second network access address of the corresponding semantic decoding model based on the semantic encoding model accessed by the original data, and sends the semantic encoded data to the second network access address.

[0099] The semantic communication server performs semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model, including: the semantic communication server performs semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the second network access address.

[0100] Preferably, the first network access address and the second network access address are URL addresses.

[0101] Preferably, the first network access address includes the IP address, port number, and corresponding semantic encoding model number of the client front-end device, and the second network access address includes the IP address, port number, and corresponding semantic decoding model number of the semantic communication server.

[0102] This invention employs the technical means of embodiments of the present invention to establish a URL-based semantic communication network architecture. On the client front-end device and semantic communication server sides, semantic encoding and decoding models are paired via URLs, enabling the deployment of multiple semantic processing models to meet the encoding and decoding needs of data with different semantic communication requirements. User terminals execute semantic communication tasks by configuring URLs, achieving adaptive semantic communication. This invention can fully utilize the computing resources of the computing network, significantly reducing the bandwidth requirements of private network users in client front-end device scenarios through semantic communication, reducing the allocation of dedicated resources for private networks, thereby greatly reducing private network costs and facilitating industry support for market promotion.

[0103] It should be noted that the semantic communication system provided in this embodiment of the invention is used to execute all the process steps of the semantic communication method applied to the semantic communication system in the above embodiment. The working principle and beneficial effect of the two correspond one-to-one, so they will not be described again.

[0104] This invention also provides a semantic communication device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the semantic communication method as described in any of the above embodiments.

[0105] This invention also provides a computer-readable storage medium, which includes a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to execute the semantic communication method as described in any of the above embodiments.

[0106] Specifically, the semantic communication device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a semantic communication program. When the processor executes the computer program, it implements the steps described in each of the semantic communication method embodiments above, for example... Figure 1 Steps S11 to S13 shown, or for example Figure 4 The steps S21-22 shown, or the processor executing the computer program, implements the functions of each module in the above-described device embodiments.

[0107] For example, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the semantic communication device. For example, the computer program can be divided into several modules, the specific functions of which have been described in detail in the semantic communication method provided in any of the above embodiments; therefore, the specific functions of the device will not be repeated here.

[0108] The semantic communication device described herein can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. The semantic communication device may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that the above embodiments are merely examples of a device and do not constitute a limitation on a semantic communication device. It may include more or fewer components than illustrated, or combine certain components, or use different components. For example, the semantic communication device may also include input / output devices, network access devices, buses, etc.

[0109] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the semantic communication device, connecting all parts of the device via various interfaces and lines.

[0110] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the semantic communication device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0111] If the module integrated into the semantic communication device is implemented as 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, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0112] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A semantic communication method, characterized in that, This method is applied to client-side devices, where user terminals connect to a semantic communication server via the client-side device; the semantic communication method includes: Receive raw data sent by the user terminal; The original data is semantically encoded using a semantic coding model to obtain semantically encoded data; The semantically encoded data is sent to the semantic communication server, so that the semantic communication server performs semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model. The different semantic encoding models of the client front-end device are used to process the semantic encoding of data with different semantic communication needs, and each of the semantic encoding models is predefined with its own first network access address; the different semantic decoding models of the semantic communication server are used to process the semantic decoding of data with different semantic communication needs, and each of the semantic decoding models is predefined with its own second network access address. The step of semantically encoding the original data using a semantic coding model to obtain semantically encoded data includes: Based on the first network access address accessed by the user terminal, the original data is semantically encoded using the semantic coding model corresponding to the first network access address to obtain semantically encoded data; wherein, the user terminal can configure the first network access address to which the original data needs to be sent according to its own semantic communication requirements, and send the original data to the first network access address. The step of sending the semantically encoded data to the semantic communication server, so that the semantic communication server performs semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model, includes: Based on the semantic encoding model accessed by the original data, a second network access address for the corresponding semantic decoding model is determined, and the semantic encoded data is sent to the second network access address so that the semantic communication server performs semantic decoding on the semantic encoded data through the semantic decoding model corresponding to the second network access address.

2. The semantic communication method as described in claim 1, characterized in that, The first network access address and the second network access address are URL addresses.

3. The semantic communication method as described in claim 2, characterized in that, The first network access address is composed of the IP address, port number, and corresponding semantic encoding model number of the client front-end device. The second network access address is composed of the IP address, port number, and corresponding semantic decoding model number of the semantic communication server.

4. The semantic communication method as described in claim 1, characterized in that, The semantic encoding model and the semantic decoding model are trained and shared by the semantic communication server.

5. The semantic communication method as described in claim 1, characterized in that, Sending the semantically encoded data to the semantic communication server includes: The semantic encoded data is channel encoded to form channel encoded data and sent to the mobile communication base station, so that the mobile communication base station performs channel decoding on the channel encoded data to obtain the semantic encoded data and sends it to the semantic communication server.

6. A semantic communication method, characterized in that, Applied to a semantic communication server, a user terminal connects to the semantic communication server through a client front-end device; different semantic encoding models of the client front-end device are used to process the semantic encoding of data with different semantic communication needs, and each semantic encoding model is predefined with its own first network access address; different semantic decoding models of the semantic communication server are used to process the semantic decoding of data with different semantic communication needs, and each semantic decoding model is predefined with its own second network access address. The semantic communication method includes: The system receives semantically encoded data sent by the client front-end device. The semantically encoded data is obtained by the client front-end device semantically encoding the original data sent by the user terminal using a semantic encoding model corresponding to the first network access address, based on the first network access address accessed by the user terminal. The user terminal can configure the first network access address to which the original data needs to be sent according to its own semantic communication requirements and send the original data to the first network access address. The client front-end device can determine the second network access address of the corresponding semantic decoding model based on the semantic encoding model accessed by the original data and send the semantically encoded data to the second network access address. The semantically encoded data is semantically decoded using the semantic decoding model corresponding to the second network access address.

7. The semantic communication method as described in claim 6, characterized in that, The semantic communication method further includes: Construct and train the semantic encoding model and the semantic decoding model; The semantic encoding model is shared to the client's front-end device for configuration.

8. The semantic communication method as described in claim 7, characterized in that, The construction and training of the semantic encoding model and the semantic decoding model include: A first neural network model for semantic encoding and a second neural network model for semantic decoding are constructed according to different semantic communication requirements, and the first neural network model is deployed to the client front-end device. Receive training data sent by the client front-end device; wherein the training data is sent from the user terminal to the client front-end device; The system receives semantic encoding training data sent by the client front-end device; wherein the semantic encoding training data is obtained by the client front-end device semantically encoding the training data through the first neural network model. The semantic encoding training data is semantically decoded using the second neural network model to obtain semantic decoding training data; The training data and the semantic decoding training data are compared, and the parameters of the first neural network model and the second neural network model are tuned according to the comparison results. Training ends when the accuracy of the first neural network model and the second neural network model reaches a preset accuracy threshold, or the number of training iterations reaches a preset number threshold. The semantic encoding model is obtained based on the first neural network model, and the semantic decoding model is obtained based on the second neural network model.

9. A semantic communication device, characterized in that, Applied to client-side devices, user terminals connect to a semantic communication server via the client-side devices; the semantic communication device includes: The raw data receiving module is used to receive raw data sent by the user terminal; The semantic encoding module is used to perform semantic encoding on the original data through a semantic encoding model to obtain semantically encoded data. The data sending module is used to send the semantically encoded data to the semantic communication server, so that the semantic communication server can perform semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the semantic encoding model; The different semantic encoding models of the client front-end device are used to process the semantic encoding of data with different semantic communication needs, and each of the semantic encoding models is predefined with its own first network access address; the different semantic decoding models of the semantic communication server are used to process the semantic decoding of data with different semantic communication needs, and each of the semantic decoding models is predefined with its own second network access address. The semantic encoding module is specifically used for: Based on the first network access address accessed by the user terminal, the original data is semantically encoded using the semantic coding model corresponding to the first network access address to obtain semantically encoded data; wherein, the user terminal can configure the first network access address to which the original data needs to be sent according to its own semantic communication requirements, and send the original data to the first network access address. The data sending module is specifically used for: Based on the semantic encoding model accessed by the original data, a second network access address for the corresponding semantic decoding model is determined, and the semantic encoded data is sent to the second network access address so that the semantic communication server performs semantic decoding on the semantic encoded data through the semantic decoding model corresponding to the second network access address.

10. A semantic communication device, characterized in that, Applied to a semantic communication server, a user terminal connects to the semantic communication server through a client front-end device; different semantic encoding models of the client front-end device are used to process the semantic encoding of data with different semantic communication needs, and each semantic encoding model is predefined with its own first network access address; different semantic decoding models of the semantic communication server are used to process the semantic decoding of data with different semantic communication needs, and each semantic decoding model is predefined with its own second network access address. The semantic communication device includes: The encoded data receiving module is used to receive semantically encoded data sent by the client front-end device. The semantically encoded data is obtained by the client front-end device semantically encoding the original data sent by the user terminal using a semantic encoding model, based on a first network access address accessed by the user terminal. The user terminal can configure the first network access address to which the original data needs to be sent according to its own semantic communication requirements and send the original data to the first network access address. The client front-end device can determine a second network access address corresponding to the semantic decoding model accessed by the original data based on the semantic encoding model and send the semantically encoded data to the second network access address. The semantic decoding module is used to perform semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the second network access address.

11. A semantic communication system, characterized in that, The semantic communication system includes a user terminal, a client front-end device, and a semantic communication server. The user terminal is connected to the semantic communication server through the client front-end device. Different semantic encoding models of the client front-end device are used to process the semantic encoding of data with different semantic communication needs, and each semantic encoding model is predefined with its own first network access address. Different semantic decoding models of the semantic communication server are used to process the semantic decoding of data with different semantic communication needs, and each semantic decoding model is predefined with its own second network access address. The user terminal is configured to configure the first network access address to which the original data needs to be sent according to its own semantic communication requirements, and send the original data to the first network access address. The client front-end device is used to semantically encode the original data through the semantic coding model corresponding to the first network access address accessed by the user terminal to obtain semantically encoded data; and to determine the second network access address of the corresponding semantic decoding model based on the semantic coding model accessed by the original data, and send the semantically encoded data to the second network access address. The semantic communication server is used to perform semantic decoding on the semantically encoded data through the semantic decoding model corresponding to the second network access address.

Images