Communication method and communication apparatus

Abstract

The present application provides a communication method and a communication apparatus. The method may comprise: acquiring a first semantic feature and first calibration information by means of an enhanced semantic encoder, wherein the first semantic feature comprises an output obtained by the enhanced semantic encoder using first data as an input, the first calibration information comprises a signal-to-noise ratio of the first semantic feature or the first data, the first data comprises noisy data, and the enhanced semantic encoder is a semantic encoder trained using noisy data; and sending the first semantic feature and the first calibration information, wherein the first semantic feature and the first calibration information are used for decoding semantic information corresponding to the first data. The method can improve the impact of noise of a sending end on the semantic understanding accuracy in a semantic communication system.

Description

Communication method and communication device

[0001] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office of China on March 22, 2024, with application number 202410350205.6 and invention name “A Communication Method and Communication Device”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of communication technology, and more particularly, to a communication method and a communication device. Background Art

[0003] The goal of semantic communication is for the receiver to correctly interpret the semantics of the message, rather than precisely reconstruct the sent data. The sender and receiver in a semantic communication system require a common semantic knowledge base. In a semantic communication system, by performing semantic extraction on the raw data at the sender and transmitting the extracted semantic information instead of the raw data, bandwidth requirements can be significantly reduced. Deep neural networks are widely used to process semantic information for various types of data, such as object classification, detection, and description in images, and natural language understanding and translation. Semantic encoders and decoders based on deep neural networks, deployed at the sender and receiver, respectively, enable semantic communication.

[0004] In a semantic communication system based on deep neural networks, the transmitter converts raw data into semantic features and sends them to the receiver, which then extracts the semantic information. During this communication process, noise on the transmitter side, such as background noise in the input speech or other user speech, can introduce semantic noise and lead to misunderstandings on the receiver side. Therefore, mitigating the impact of transmitter noise on semantic understanding accuracy in semantic communication systems is an urgent challenge in this field. Summary of the Invention

[0005] The present application provides a communication method and a communication device, which can improve the impact of noise at a transmitting end on the accuracy of semantic information extracted at a receiving end.

[0006] In a first aspect, a communication method is provided, the method comprising: obtaining a first semantic feature and first calibration information through an enhanced semantic encoder, the first semantic feature comprising an output obtained by the enhanced semantic encoder with first data as input, the first calibration information comprising the first semantic feature or a signal-to-noise ratio of the first data, the first data comprising noisy data, and the enhanced semantic encoder being a semantic encoder trained using the noisy data; sending the first semantic feature and the first calibration information, the first semantic feature and the first calibration information being used to decode semantic information corresponding to the first data.

[0007] Specifically, the first calibration information may include a signal-to-noise ratio of the first data or the first semantic feature. For example, the first calibration information may include a noise level or a distortion position of the first semantic feature or the first data.

[0008] Specifically, the first calibration information can be directly output by the enhanced semantic encoder, or the first calibration information can also be obtained by processing the first data or the first semantic feature based on another neural network, which can be trained based on the performance of the enhanced semantic encoder or the enhanced semantic decoder, or the difference between the received semantic feature and the sent semantic feature.

[0009] In an embodiment of the present application, a first device sends a first semantic feature and first calibration information, wherein the first semantic feature and the first calibration information may be information obtained after the enhanced semantic encoder in the first device inputs noisy data, and the enhanced semantic encoder may be an encoder independently trained with noisy data. The second device that receives the first semantic feature and the first calibration information may use the first calibration information to assist the decoder in decoding the first semantic feature, thereby increasing the reliability of the decoder in the second device, so that the semantic communication system can more accurately obtain the semantic information corresponding to the first data when the input first data contains noise.

[0010] In combination with the first aspect, in some implementations of the first aspect, before obtaining the first semantic feature and the first calibration information through the enhanced semantic encoder, it also includes: obtaining the enhanced semantic encoder; obtaining the second semantic feature and the second calibration information through the enhanced semantic encoder, the second semantic feature includes the output obtained by the enhanced semantic encoder with the second data as input, the second calibration information includes the second semantic feature or the signal-to-noise ratio of the second data, and the second data includes noisy data; sending the second semantic feature, the second calibration information and the first semantic information, the first semantic information includes a label corresponding to the second data, and the second semantic feature, the second calibration information and the first semantic information are used to adjust the basic semantic decoder to obtain the enhanced semantic decoder.

[0011] Specifically, the second calibration information may include the output obtained after the enhanced semantic encoder inputs the noisy second data, or the second calibration information may include the output obtained by processing the input data or output semantic features of the encoder based on another neural network. The second calibration information can be used together with the second semantic features as the input of the basic semantic decoder in the second device, and is used to adjust the basic semantic decoder to obtain an enhanced semantic decoder.

[0012] In an embodiment of the present application, the enhanced semantic encoder can be an encoder independently trained with noisy data. The first device sends the first semantic information and the second semantic features and second calibration information obtained by the enhanced semantic encoder using the second data as input, so that the second device that receives the second semantic features, the second calibration information and the first semantic information can use the calibration information to assist the second semantic features and the first semantic information to adjust the basic semantic decoder to obtain an enhanced semantic decoder, so that the semantic communication system can use the second calibration information to train a more reliable decoder.

[0013] In combination with the first aspect, in certain implementations of the first aspect, the first calibration information and / or the second calibration information are processed using a cyclic redundancy check CRC and / or a retransmission mechanism when being sent.

[0014] In an embodiment of the present application, the calibration information can be transmitted from the first device to the second device in a more reliable mode by adopting a cyclic redundancy check CRC and / or a retransmission mechanism, thereby ensuring the accuracy of the calibration information when used for auxiliary decoding and / or for auxiliary training of the decoder, and improving the performance of reasoning and training of the semantic communication system.

[0015] In combination with the first aspect, in some implementations of the first aspect, before sending the second semantic feature, the second calibration information and the first semantic information, it also includes: receiving first indication information, the first indication information is used to request the second semantic feature, the second calibration information and the first semantic information.

[0016] In an embodiment of the present application, the first device specifically determines whether to send the second semantic feature, the second calibration information and the first semantic information based on the first indication information sent by the second device. Since the second semantic feature, the second calibration information and the first semantic information are used to adjust the basic semantic decoder to obtain an enhanced semantic decoder, the decoder in the second device can determine the first indication information based on its own performance, and inform the first device decoder whether it needs to be adjusted through the first indication information. If the decoder needs to be adjusted, the first device including the encoder will send the second semantic feature, the second calibration information and the first semantic information to the second device, so that the adjustment of the decoder in the semantic communication system is more flexible.

[0017] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: sending a third semantic feature and a fourth semantic feature, the third semantic feature and the fourth semantic feature are used to determine the first indication information, the third semantic feature includes the output of the enhanced semantic encoder with the third data as input, and the fourth semantic feature includes the output of the enhanced semantic encoder with the noisy third data as input.

[0018] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: sending a fourth semantic feature and second semantic information, the fourth semantic feature and the second semantic information being used to determine the first indication information, the fourth semantic feature including the output of the enhanced semantic encoder with the noisy third data as input, and the second semantic information including a label corresponding to the third data.

[0019] In an embodiment of the present application, the first indication information sent by the second device to the first device is determined based on the third semantic feature and the fourth semantic feature or based on the four semantic features and the second semantic information, and the second semantic feature, the second calibration information and the first semantic information specifically requested by the first indication information are determined based on the content of the first indication information, so that the decoder can flexibly select training data to achieve adjustment.

[0020] In combination with the first aspect, in certain implementations of the first aspect, the method also includes: sending multiple calibration information types, the first indication information is also used to indicate that the second calibration information is the first type of calibration information among the multiple calibration information types, the second semantic feature includes a semantic feature associated with the first type of calibration information, and the first semantic information includes semantic information associated with the first type of calibration information.

[0021] Specifically, when the distance between the noisy enhanced semantic features and the noiseless enhanced semantic feature elements of a batch of data is evenly distributed, it can be judged that the semantic noise in the batch of data or the noisy enhanced semantic features corresponding to the batch of data is evenly distributed, and the calibration information can be used to indicate the noise intensity of each element or the average noise intensity in the noisy enhanced semantic features. The calibration information obtained through the batch of data corresponds to a type with higher accuracy. For example, the calibration information obtained through the batch of data corresponds to a type close to type 1 in Figure 10.

[0022] Specifically, when the distance distribution between the noisy enhanced semantic features and the noiseless enhanced semantic feature elements of a batch of data is uneven, for example, the distance distribution is that the distance between some noisy enhanced semantic features and the noiseless enhanced semantic feature elements is large, and the distance between some noisy enhanced semantic features and the noiseless enhanced semantic feature elements is small, it can be judged as burst semantic noise, and the calibration information can be used to indicate the semantic feature 0 / 1 mask and noise intensity, for example, assigning a weight of 1 to the semantic feature with large noise and a weight of 0 to the semantic feature with small noise. The calibration information obtained through this batch of data corresponds to a type with lower accuracy. For example, the calibration information obtained through this batch of data corresponds to a type close to type K in Figure 10.

[0023] In an embodiment of the present application, a first device sends multiple calibration information types to a second device. The second device can specifically select the first type from the multiple calibration information types based on its own performance, and inform the first device of the first type through a first indication message. Then, the first device can only send the first type of calibration information, as well as semantic features and semantic information tags associated with the first type of calibration information, to the second device for adjustment of the decoder in the second device. It can be seen that the second device can select some data with better effects to adjust the decoder based on its own performance, thereby reducing the time and information transmission resources required for decoder adjustment and increasing the adjustment efficiency of the decoder.

[0024] In combination with the first aspect, in certain implementations of the first aspect, obtaining an enhanced semantic encoder includes: training a first neural network, wherein the noisy fifth data is the input of the first neural network, the fifth semantic feature is the target output of the first neural network, the fifth semantic feature includes the output of a basic semantic encoder with the fifth data as input, the basic semantic encoder is a pre-trained encoder, and the enhanced semantic encoder includes the trained first neural network.

[0025] In an embodiment of the present application, the enhanced semantic encoder can be obtained using input data and a basic semantic encoder, that is, the acquisition of the enhanced semantic encoder can be completed independently by the encoder end. Compared with the semantic communication system that overcomes noise through joint training, the independently trained encoder can stagger the time of encoder training and decoder training, and flexibly select the time to transmit the data required for training the decoder. In addition, the amount of data used for encoder training and decoder training can also be different, which increases the flexibility of semantic communication system training.

[0026] In combination with the first aspect, in certain implementations of the first aspect, when the distance between the sixth semantic feature and the seventh semantic feature is greater than or equal to a first threshold, an enhanced semantic encoder is obtained; the sixth semantic feature includes the output of the basic semantic encoder with the sixth data as input, the seventh semantic feature includes the output of the basic semantic encoder with the noisy sixth data as input, and the basic semantic encoder is a pre-trained encoder.

[0027] In an embodiment of the present application, whether the first device obtains the enhanced semantic encoder is determined based on the distance between the semantic features output when the basic semantic encoder in the first device inputs noisy data and noise-free data. That is, the first device can reasonably judge whether it is necessary to train the enhanced semantic encoder based on the performance of the basic semantic encoder in processing noise, thereby increasing the flexibility of training the semantic communication system.

[0028] In a second aspect, a communication method is provided, which includes: receiving a first semantic feature and first calibration information, the first semantic feature and the first calibration information being obtained through an enhanced semantic encoder, the first semantic feature including the output obtained by the enhanced semantic encoder with first data as input, the first calibration information including the first semantic feature or the signal-to-noise ratio of the first data, the first data including noisy data, and the enhanced semantic encoder being a semantic encoder trained with the noisy data; and using the first semantic feature and the first calibration information to decode semantic information corresponding to the first data.

[0029] In combination with the second aspect, in some implementations of the second aspect, before receiving the second semantic feature, the second calibration information and the first semantic information, it also includes: receiving the second semantic feature, the second calibration information and the first semantic information, the second semantic feature includes the output of the enhanced semantic encoder with the second data as input, the second calibration information includes the second semantic feature or the signal-to-noise ratio of the second data, the second data includes noisy data, and the first semantic information includes a label corresponding to the second data; using the second semantic feature, the second calibration information and the first semantic information to adjust the basic semantic decoder to obtain an enhanced semantic decoder.

[0030] In combination with the second aspect, in certain implementations of the second aspect, the first calibration information and / or the second calibration information are processed using a cyclic redundancy check CRC and / or a retransmission mechanism when being sent.

[0031] In combination with the second aspect, in some implementations of the second aspect, before receiving the second semantic feature, the second calibration information and the first semantic information, it also includes: sending first indication information, the first indication information is used to request the second semantic feature, the second calibration information and the first semantic information.

[0032] In combination with the second aspect, in certain implementations of the second aspect, the method further includes: receiving a third semantic feature and a fourth semantic feature, the third semantic feature and the fourth semantic feature are used to determine the first indication information, the third semantic feature includes the output obtained by the enhanced semantic encoder with the third data as input, and the fourth semantic feature includes the output obtained by the enhanced semantic encoder with the noisy third data as input.

[0033] In combination with the second aspect, in certain implementations of the second aspect, the method also includes: when the first distance between the third semantic information and the fourth semantic information is greater than or equal to a second threshold, it is determined that the basic semantic decoder needs to be adjusted, the third semantic information includes the output obtained by the basic semantic decoder with the third semantic feature as input, the fourth semantic information includes the output obtained by the basic semantic decoder with the fourth semantic feature as input, and the first distance includes the cosine distance or Euclidean distance between the third semantic information and the fourth semantic information.

[0034] In combination with the second aspect, in certain implementations of the second aspect, the method further includes: receiving a fourth semantic feature and second semantic information, the fourth semantic feature and the second semantic information being used to determine the first indication information, the fourth semantic feature including the output of the enhanced semantic encoder with the noisy third data as input, and the second semantic information including a label corresponding to the third data.

[0035] In combination with the second aspect, in certain implementations of the second aspect, the method also includes: when the first distance between the second semantic information and the fourth semantic information is greater than or equal to a second threshold, determining that the basic semantic decoder needs to be adjusted, the fourth semantic information includes the output obtained by the basic semantic decoder with the fourth semantic feature as input, and the first distance includes the cosine distance or Euclidean distance between the second semantic information and the fourth semantic information.

[0036] In an embodiment of the present application, whether the decoder needs to be adjusted can be judged based on the distance between the third semantic information and the fourth semantic information or the distance between the second semantic information and the fourth semantic information. That is, the decoder can flexibly choose whether to make adjustments based on its own performance in processing noise, thereby improving the flexibility of semantic communication system training.

[0037] In combination with the second aspect, in certain implementations of the second aspect, the method also includes: receiving multiple calibration information types, the first indication information is also used to indicate that the second calibration information is the first type of calibration information among the multiple calibration information types, the second semantic feature includes a semantic feature associated with the first type of calibration information, and the first semantic information includes semantic information associated with the first type of calibration information.

[0038] In combination with the second aspect, in some implementations of the second aspect, the first type is determined according to the first distance.

[0039] In combination with the second aspect, in certain implementations of the second aspect, the basic semantic decoder is adjusted using the second semantic feature, the second calibration information, and the first semantic information to obtain an enhanced semantic decoder, including: training the basic semantic decoder, wherein the second semantic feature and the second calibration information are inputs to the basic semantic decoder, the first semantic information is the target output of the basic semantic decoder training, and the second semantic feature, the second calibration information, and the first semantic information are used to adjust the weights of the basic semantic decoder neural network.

[0040] Specifically, the second calibration information can be cascaded with the second semantic feature as a training input for adjusting the basic semantic decoder to obtain an enhanced semantic decoder, or as an input for the enhanced semantic decoder during semantic reasoning, and the first semantic information can be used as the target output of decoder training to adjust the weights of the decoder neural network.

[0041] Specifically, the second calibration information can be decoded by a second neural network assisted decoder, the input of the second neural network is the second calibration information, the output of the second neural network is associated with a partial layer of the decoder neural network, and the second neural network is used to adjust the weights of the partial layer of the decoder through the calibration information.

[0042] Specifically, the second calibration information can be decoded by a third neural network assisted decoder, the input of the third neural network is the second semantic feature and the second calibration information, the output of the third neural network is the input of the decoder neural network, and the third neural network is used to preprocess the second semantic feature based on the second calibration information.

[0043] Specifically, the second calibration information is processed by a neural network, and in the process of assisting the enhanced semantic decoder in decoding, it can refer to fine-tuning some neural network parameters of the decoder, for example, the output of the second neural network or the third neural network is connected to some layers of the decoding neural network to adjust the weights of some layers.

[0044] In an embodiment of the present application, the decoder can use the second semantic feature, the second calibration information and the first semantic information to adjust the weight of the basic semantic decoder during training, and the second calibration information can be used to increase the reliability of the adjusted decoder.

[0045] According to a third aspect, a communication device is provided, which includes: a processing unit, the processing unit is used to obtain a first semantic feature and first calibration information through an enhanced semantic encoder, the first semantic feature includes the output of the enhanced semantic encoder with first data as input, the first calibration information includes the first semantic feature or the signal-to-noise ratio of the first data, the first data includes noisy data, and the enhanced semantic encoder is a semantic encoder trained with the noisy data; the device also includes: a transceiver unit, the transceiver unit is used to send the first semantic feature and the first calibration information, the first semantic feature and the first calibration information are used to decode semantic information corresponding to the first data.

[0046] In combination with the third aspect, in certain implementations of the third aspect, the processing unit is further used to obtain an enhanced semantic encoder; the processing unit is further used to obtain a second semantic feature and second calibration information through the enhanced semantic encoder, the second semantic feature includes the output obtained by the enhanced semantic encoder with the second data as input, the second calibration information includes the second semantic feature or the signal-to-noise ratio of the second data, and the second data includes noisy data; the transceiver unit is further used to send the second semantic feature, the second calibration information and the first semantic information, the first semantic information includes a label corresponding to the second data, and the second calibration information is used to assist the second semantic feature and the first semantic information in adjusting the basic semantic decoder to obtain an enhanced semantic decoder.

[0047] In combination with the third aspect, in certain implementations of the third aspect, the first calibration information and / or the second calibration information are processed using a cyclic redundancy check CRC and / or a retransmission mechanism when being sent.

[0048] In combination with the third aspect, in some implementations of the third aspect, the transceiver unit is further used to receive first indication information, where the first indication information is used to request the second semantic feature, the second calibration information, and the first semantic information.

[0049] In combination with the third aspect, in certain implementations of the third aspect, the transceiver unit is further used to send a third semantic feature and a fourth semantic feature, and the third semantic feature and the fourth semantic feature are used to determine the first indication information. The third semantic feature includes the output of the enhanced semantic encoder with the third data as input, and the fourth semantic feature includes the output of the enhanced semantic encoder with the noisy third data as input.

[0050] In combination with the third aspect, in certain implementations of the third aspect, the transceiver unit is also used to send a fourth semantic feature and second semantic information, the fourth semantic feature and the second semantic information are used to determine the first indication information, the fourth semantic feature includes the output of the enhanced semantic encoder with noisy third data as input, and the second semantic information includes a label corresponding to the third data.

[0051] In combination with the third aspect, in certain implementations of the third aspect, the transceiver unit is also used to send multiple calibration information types, the first indication information is also used to indicate that the second calibration information is the first type of calibration information among multiple calibration information types, the second semantic feature includes a semantic feature associated with the first type of calibration information, and the first semantic information includes semantic information associated with the first type of calibration information.

[0052] In combination with the third aspect, in certain implementations of the third aspect, the processing unit is further used to train the first neural network, wherein the noisy fifth data is the input of the first neural network, the fifth semantic feature is the target output of the first neural network, the fifth semantic feature includes the output of the basic semantic encoder with the fifth data as input, the basic semantic encoder is a pre-trained encoder, and the enhanced semantic encoder includes the trained first neural network.

[0053] In combination with the third aspect, in certain implementations of the third aspect, when the distance between the sixth semantic feature and the seventh semantic feature is greater than or equal to a first threshold, an enhanced semantic encoder is obtained; the sixth semantic feature includes the output of the basic semantic encoder with the sixth data as input, the seventh semantic feature includes the output of the basic semantic encoder with the noisy sixth data as input, and the basic semantic encoder is a pre-trained encoder.

[0054] In a fourth aspect, a communication device is provided, which includes: a transceiver unit, the transceiver unit is used to receive a first semantic feature and first calibration information, the first semantic feature and the first calibration information are obtained through an enhanced semantic encoder, the first semantic feature includes the output obtained by the enhanced semantic encoder with the first data as input, the first calibration information includes the first semantic feature or the signal-to-noise ratio of the first data, the first data includes noisy data, and the enhanced semantic encoder is a semantic encoder trained with the noisy data; the device also includes: a processing unit, the processing unit is used to decode the semantic information corresponding to the first data based on the first semantic feature and the first calibration information.

[0055] In combination with the fourth aspect, in certain implementations of the fourth aspect, the transceiver unit is further used to receive a second semantic feature, second calibration information, and first semantic information, the second semantic feature including the output of the enhanced semantic encoder with the second data as input, the second calibration information including the second semantic feature or the signal-to-noise ratio of the second data, the second data including noisy data, and the first semantic information including a label corresponding to the second data; the processing unit is further used to use the second calibration information to assist the second semantic feature and the first semantic information in adjusting the basic semantic decoder to obtain an enhanced semantic decoder.

[0056] In combination with the fourth aspect, in certain implementations of the fourth aspect, the first calibration information and / or the second calibration information are processed using a cyclic redundancy check CRC and / or a retransmission mechanism when being sent.

[0057] In combination with the fourth aspect, in some implementations of the fourth aspect, the transceiver unit is further used to send first indication information, where the first indication information is used to request the second semantic feature, the second calibration information, and the first semantic information.

[0058] In combination with the fourth aspect, in certain implementations of the fourth aspect, the transceiver unit is further used to receive a third semantic feature and a fourth semantic feature, the third semantic feature and the fourth semantic feature are used to determine the first indication information, the third semantic feature includes the output of the enhanced semantic encoder with the third data as input, and the fourth semantic feature includes the output of the enhanced semantic encoder with the noisy third data as input.

[0059] In combination with the fourth aspect, in certain implementations of the fourth aspect, when the first distance between the third semantic information and the fourth semantic information is greater than or equal to a second threshold, it is determined that the basic semantic decoder needs to be adjusted, the third semantic information includes the output obtained by the basic semantic decoder with the third semantic feature as input, the fourth semantic information includes the output obtained by the basic semantic decoder with the fourth semantic feature as input, and the first distance includes the cosine distance or Euclidean distance between the third semantic information and the fourth semantic information.

[0060] In combination with the fourth aspect, in certain implementations of the fourth aspect, the transceiver unit is further used to receive a fourth semantic feature and second semantic information, the fourth semantic feature and the second semantic information are used to determine the first indication information, the fourth semantic feature includes the output of the enhanced semantic encoder with noisy third data as input, and the second semantic information includes a label corresponding to the third data.

[0061] In combination with the fourth aspect, in certain implementations of the fourth aspect, when the first distance between the second semantic information and the fourth semantic information is greater than or equal to a second threshold, it is determined that the basic semantic decoder needs to be adjusted, the fourth semantic information includes the output obtained by the basic semantic decoder with the fourth semantic feature as input, and the first distance includes the cosine distance or Euclidean distance between the second semantic information and the fourth semantic information.

[0062] In combination with the fourth aspect, in certain implementations of the fourth aspect, the transceiver unit is also used to receive multiple calibration information types, the first indication information is also used to indicate that the second calibration information is the first type of calibration information among the multiple calibration information types, the second semantic feature includes a semantic feature associated with the first type of calibration information, and the first semantic information includes semantic information associated with the first type of calibration information.

[0063] In combination with the fourth aspect, in certain implementations of the fourth aspect, the first type is determined according to the first distance.

[0064] In combination with the fourth aspect, in certain implementations of the fourth aspect, the processing unit is also used to train a basic semantic decoder, wherein the second semantic feature is the input of the basic semantic decoder, the first semantic information is the target output of the basic semantic decoder, and the second calibration information is used to adjust the weights of some neural network layers of the basic semantic decoder.

[0065] In a fifth aspect, a communication device is provided, comprising: a processor coupled to a memory, the memory being used to store a computer program, the processor being used to run the computer program, so that the communication device executes the method as described in the first aspect and any possible implementation thereof.

[0066] In a sixth aspect, a communication device is provided, comprising: a processor coupled to a memory, the memory being used to store a computer program, the processor being used to run the computer program, so that the communication device executes the method as described in the second aspect and any possible implementation thereof.

[0067] In the seventh aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a computer, the computer executes the communication method that can be implemented in the first aspect and the first aspect, or the second aspect and the second aspect.

[0068] In an eighth aspect, a computer program product comprising instructions is provided, which, when executed on a computer, enables the computer to execute any of the communication methods that can be implemented in the first aspect and the first aspect, or the second aspect and the second aspect.

[0069] In the ninth aspect, a chip is provided, which includes a processor and a data interface. The processor reads instructions stored in a memory through the data interface to execute any communication method that can be implemented in the first aspect and the first aspect, or the second aspect and the second aspect.

[0070] In combination with the ninth aspect, in one possible implementation, the processor is coupled to the memory through an interface.

[0071] In combination with the ninth aspect, in one possible implementation, the chip system also includes a memory, in which a computer program or computer instructions are stored. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] FIG1 is a schematic diagram of a wireless communication system applicable to an embodiment of the present application.

[0073] Figure 2 is a schematic diagram of the neuron structure.

[0074] FIG3 is a schematic diagram of a semantic communication system based on a neural network.

[0075] FIG4 is a schematic diagram of a semantic communication system with semantic noise.

[0076] FIG5 is a flow chart of a semantic communication system training and application method 500 provided in an embodiment of the present application.

[0077] FIG6 is a schematic diagram of an enhanced semantic encoder training provided in an embodiment of the present application.

[0078] FIG7 is a decoding diagram of an enhanced semantic decoder provided in an embodiment of the present application.

[0079] FIG8 is a semantic communication system module provided in an embodiment of the present application.

[0080] FIG9 is a flow chart of another semantic communication system training and application method 900 provided in an embodiment of the present application.

[0081] FIG10 is a schematic diagram of obtaining a calibration information type provided in an embodiment of the present application.

[0082] FIG11 is a schematic diagram of calibration information type selection provided in an embodiment of the present application.

[0083] FIG12 is a schematic diagram of a functional module of a semantic communication system chip provided in an embodiment of the present application.

[0084] FIG13 is a schematic structural diagram of a communication device provided in an embodiment of the present application.

[0085] FIG14 is a schematic diagram of a communication architecture provided in an embodiment of the present application.

[0086] FIG15 is a schematic diagram of a chip system 1500 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0087] The technical solution in this application will be described below with reference to the accompanying drawings.

[0088] First, a brief introduction is given to the communication system to which this application is applicable.

[0089] The technical solutions provided in this application can be applied to various communication systems, such as: fifth generation (5G) or new radio (NR) systems, long term evolution (LTE) systems, LTE frequency division duplex (FDD) systems, LTE time division duplex (TDD) systems, wireless local area networks (WLAN) systems, satellite communication systems, future communication systems, such as sixth generation (6G) mobile communication systems, or a fusion system of multiple systems. The technical solutions provided in this application can also be applied to device to device (D2D) communication, vehicle-to-everything (V2X) communication, machine to machine (M2M) communication, machine type communication (MTC), and Internet of Things (IoT) communication systems or other communication systems.

[0090] A device in a communication system can send signals to or receive signals from another device. These signals may include information, signaling, or data. The term "device" can also be replaced by an entity, network entity, network element, communication device, communication module, node, communication node, and the like. This disclosure uses devices as examples for description. For example, a communication system may include at least one terminal device and at least one network device. A network device can send downlink signals to a terminal device, and / or a terminal device can send uplink signals to a network device.

[0091] The terminal devices in the embodiments of the present application include various devices with wireless communication functions, which can be used to connect people, objects, machines, etc. The terminal devices can be widely used in various scenarios, such as: cellular communication, D2D, V2X, peer to peer (P2P), M2M, MTC, IoT, virtual reality (VR), augmented reality (AR), industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearable, smart transportation, smart city drones, robots, remote sensing, passive sensing, positioning, navigation and tracking, autonomous delivery, etc. The terminal device can be a terminal in any of the above scenarios, such as an MTC terminal, an IoT terminal, etc. The terminal device may be a user equipment (UE) of the third generation partnership project (3GPP) standard, a terminal, a fixed device, a mobile station device or a mobile device, a subscriber unit, a handheld device, a vehicle-mounted device, a wearable device, a cellular phone, a smart phone, a SIP phone, a wireless data card, a personal digital assistant (PDA), a computer, a tablet computer, a notebook computer, a wireless modem, a handheld device, a laptop computer, a computer with wireless transceiver function, a smart book, a vehicle, a satellite, a global positioning system (GPS) device, a target tracking device, an aircraft (such as a drone, a helicopter, a multi-copter, a quadcopter, or an airplane), a ship, a remote control device, a smart home device, an industrial device, or a device built into the above-mentioned device (such as a communication module, a modem or a chip in the above-mentioned device), or other processing devices connected to a wireless modem. For ease of description, the terminal device will be described below by taking the terminal or UE as an example.

[0092] It should be understood that in some scenarios, a UE can also be used to act as a base station. For example, a UE can act as a scheduling entity that provides sidelink signals between UEs in scenarios such as V2X, D2D, or P2P.

[0093] In the embodiments of the present application, the device for implementing the function of the terminal device can be the terminal device, or it can be a device that can support the terminal device to implement the function, such as a chip system or chip, which can be installed in the terminal device. In the embodiments of the present application, the chip system can be composed of chips, or it can include chips and other discrete devices. In the embodiments of the present application, only the terminal device is used as an example for description, and the embodiments of the present application are not limited to the solutions of the embodiments of the present application.

[0094] The network device in the embodiments of the present application may be a device for communicating with a terminal device, and may also be referred to as an access network device or a radio access network device. For example, the network device may be a base station. The network device in the embodiments of the present application may refer to a radio access network (RAN) node (or device) that connects a terminal device to a wireless network. Base station can broadly cover various names as follows, or replace the following names, such as: NodeB, evolved NodeB (eNB), next generation NodeB (gNB), relay station, access point, transmission point (TRP), transmission point (TP), master station, auxiliary station, multi-standard wireless (motor slide retainer, MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), positioning node, etc. The base station can be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. The base station can also refer to a communication module, modem or chip used to be set in the aforementioned device or apparatus. The base station can also be a mobile switching center and a device that performs base station functions in D2D, V2X, and M2M communications, a network-side device in a 6G network, or a device that performs base station functions in future communication systems. The base station can support networks with the same or different access technologies. The embodiments of this application do not limit the specific technology and specific device form used by the network equipment.

[0095] Base stations can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and at least one cell can move based on the location of the mobile base station. In other examples, a helicopter or drone can be configured to act as a device that communicates with another base station.

[0096] In some deployments, the network devices mentioned in the embodiments of the present application may include a CU, a DU, or both a CU and a DU, or a control plane CU node (central unit-control plane (CU-CP)), a user plane CU node (central unit-user plane (CU-UP)), and a DU node. For example, the network devices may include a gNB-CU-CP, a gNB-CU-UP, and a gNB-DU.

[0097] In some deployments, multiple RAN nodes collaborate to assist terminals in achieving wireless access, with different RAN nodes implementing portions of the base station's functionality. For example, a RAN node can be a CU, DU, CU-CP, CU-UP, or RU. The CU and DU can be separate or included in the same network element, such as the BBU. The RU can be included in a radio frequency device or radio unit, such as an RRU, AAU, or RRH.

[0098] The RAN node can support one or more types of fronthaul interfaces. Different fronthaul interfaces correspond to DUs and RUs with different functions. If the fronthaul interface between the DU and the RU is the common public radio interface (CPRI), the DU is configured to implement one or more baseband functions, and the RU is configured to implement one or more radio frequency functions. If the fronthaul interface between the DU and the RU is the enhanced common public radio interface (eCPRI), relative to CPRI, part of the downlink and / or uplink baseband functions are moved from the DU to the RU for implementation. The division between the DU and the RU is different, corresponding to different types (category, Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, and F.

[0099] Taking eCPRI Cat A as an example, for downlink transmission, based on layer mapping, the DU is configured to implement layer mapping and one or more functions preceding it (i.e., one or more of coding, rate matching, scrambling, modulation, and layer mapping). Other functions after layer mapping (e.g., RE mapping, digital beamforming (BF), or one or more of inverse fast Fourier transform (IFFT) / cyclic prefix (CP) addition) are moved to the RU for implementation. For uplink transmission, based on RE demapping, the DU is configured to implement demapping and one or more functions preceding it (i.e., one or more of decoding, rate matching, descrambling, demodulation, inverse discrete Fourier transform (IDFT), channel equalization, and RE demapping). Other functions after demapping (e.g., one or more of digital BF or fast Fourier transform (FFT) / CP removal) are moved to the RU for implementation. It is understandable that for the functional description of DU and RU corresponding to various types of eCPRI, reference can be made to the eCPRI protocol, which will not be described in detail here.

[0100] In one possible design, the processing unit for implementing baseband functions in the BBU is called a baseband high layer (BBH) unit, and the processing unit for implementing baseband functions in the RRU / AAU / RRH is called a baseband low layer (BBL) unit.

[0101] In different systems, CU (or CU-CP and CU-UP), DU or RU may also have different names, but those skilled in the art can understand their meanings. For example, in the ORAN system, CU may also be called O-CU (Open CU), DU may also be called O-DU, CU-CP may also be called O-CU-CP, CU-UP may also be called O-CU-UP, and RU may also be called O-RU. Any unit of CU (or CU-CP, CU-UP), DU and RU in this application can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0102] In the embodiments of the present application, the device for implementing the function of the network device can be a network device, or a device that can support the network device to implement the function, such as a chip system or chip, which can be installed in the network device. In the embodiments of the present application, the chip system can be composed of chips, or it can include chips and other discrete devices. In the embodiments of the present application, only the device for implementing the function of the network device is a network device as an example for description, and does not constitute a limitation on the solutions of the embodiments of the present application.

[0103] Network devices and terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on the water surface; they can also be deployed on aircraft, balloons and satellites in the air. The embodiments of this application do not limit the scenarios in which network devices and terminal devices are located. In addition, terminal devices and network devices can be hardware devices, or they can be software functions running on dedicated hardware, software functions running on general-purpose hardware, such as virtualization functions instantiated on a platform (e.g., a cloud platform), or entities including dedicated or general-purpose hardware devices and software functions. This application does not limit the specific forms of terminal devices and network devices.

[0104] In addition, in order to support AI technology in wireless networks, AI nodes may also be introduced into the network.

[0105] Optionally, the AI ​​node can be deployed in one or more of the following locations in the communication system: access network equipment, terminal equipment, or core network equipment. Alternatively, the AI ​​node can be deployed separately, for example, in a location other than any of the above devices, such as a host or cloud server in an over-the-top (OTT) system. The AI ​​node can communicate with other devices in the communication system, such as one or more of the following: network equipment, terminal equipment, or core network elements.

[0106] It is understood that this application does not limit the number of AI nodes. For example, when there are multiple AI nodes, the multiple AI nodes can be divided based on function, such as different AI nodes are responsible for different functions.

[0107] It can also be understood that AI nodes can be independent devices, or they can be integrated into the same device to implement different functions, or they can be network elements in hardware devices, or they can be software functions running on dedicated hardware, or they can be virtualized functions instantiated on a platform (for example, a cloud platform). This application does not limit the specific form of the above-mentioned AI nodes.

[0108] FIG1 is a schematic diagram of a wireless communication system applicable to an embodiment of the present application.

[0109] As shown in Figure 1, the wireless communication system includes a wireless access network 100. The wireless access network 100 can be a next-generation (e.g., 6G or higher) wireless access network, or a traditional (e.g., 5G, 4G, 3G, or 2G) wireless access network. One or more terminal devices (120a-120j, collectively referred to as 120) can be connected to each other or to one or more network devices (110a, 110b, collectively referred to as 110) in the wireless access network 100. Network elements in the wireless communication system are connected through interfaces (e.g., NG, Xn) or air interfaces. In addition, one or more AI modules may be provided in each network element in the wireless communication system. The AI ​​modules deployed in different network elements may be the same or different.

[0110] FIG1 is only a schematic diagram. The wireless communication system may further include other devices, such as core network devices, wireless relay devices and / or wireless backhaul devices, which are not shown in FIG1 .

[0111] To facilitate understanding of the embodiments of the present application, the following briefly describes the relevant concepts and technologies involved in the present application.

[0112] 1. Artificial Intelligence: This refers to the ability of machines to learn, accumulate experience, and solve problems that humans can solve through experience, such as natural language understanding, image recognition, and chess. Artificial Intelligence can be understood as the intelligence exhibited by machines created by humans. Generally, AI refers to the technology that represents human intelligence through computer programs. The goals of AI include understanding intelligence by constructing computer programs that can perform symbolic reasoning or deduction.

[0113] 2. Machine learning: This is an implementation of artificial intelligence. Machine learning is a method that empowers machines to learn, enabling them to perform functions that cannot be accomplished through direct programming. In practical terms, machine learning utilizes data to train models and then uses these models to make predictions. There are many machine learning methods, such as neural networks (NNs), decision trees, and support vector machines. Machine learning theory primarily involves the design and analysis of algorithms that enable computers to learn automatically. Machine learning algorithms automatically analyze data to identify patterns and use these patterns to make predictions about unknown data.

[0114] 3. Neural Network: A specific embodiment of machine learning. A neural network is a mathematical model that processes information by mimicking the behavioral characteristics of animal neural networks. The concept of a neural network is derived from the neuronal structure of the brain. Each neuron performs a weighted sum operation on its input values, and the result of this weighted summation is passed through an activation function to generate an output.

[0115] Figure 2 is a schematic diagram of the neuron structure. As shown in Figure 2, assume that the input of the neuron is x = [x0, x1, ..., x n ], and the weights corresponding to each input are w=[w,w1,…,w n ], the weighted summation bias is b. Where b can be an integer, a decimal, or a complex number, etc. The activation function can be diversified. As an example, assuming that the activation function of a neuron is: y = f(z) = max(0,z), then the output of the neuron is: As another example, suppose the activation function of a neuron is: y = f(z) = z, then the output of the neuron is: As shown in Figure 2. The activation functions of different neurons in a neural network can be the same or different.

[0116] Neural networks generally comprise a multi-layer structure, with each layer comprising one or more logical decision units, referred to as neurons. Increasing the depth and / or width of a neural network can enhance its expressive power, providing more powerful information extraction and abstract modeling capabilities for complex systems. The depth of a neural network can be understood as the number of layers it comprises, and the number of neurons in each layer can be referred to as the width of that layer. In one possible implementation, a neural network comprises an input layer and an output layer. The input layer of the neural network processes the input through neurons and then passes the result to the output layer, which then obtains the output of the neural network. In another possible implementation, a neural network comprises an input layer, a hidden layer, and an output layer. The input layer of the neural network processes the input through neurons and then passes the result to an intermediate hidden layer. The hidden layer then passes the calculation result to the output layer or an adjacent hidden layer, which then obtains the output of the neural network. A neural network can comprise one or more sequentially connected hidden layers, without limitation.

[0117] 4. Semantic Communication: The goal of semantic communication is for the receiver to correctly interpret the semantics of the message, rather than precisely reconstruct the sent data. Therefore, the sender and receiver need a common semantic knowledge base. By extracting semantic information from the original data at the sender and transmitting the extracted semantic information instead of the original data, bandwidth requirements can be significantly reduced.

[0118] Figure 3 is a schematic diagram of a neural network-based semantic communication system. Neural networks are widely used for semantic information processing across various types of data, such as object classification, detection, and description in images, and natural language understanding and translation. A semantic encoder and decoder based on a deep neural network, deployed at the sending and receiving ends, respectively, enable semantic communication.

[0119] Figure 4 is a schematic diagram of a semantic communication system with semantic noise. In a semantic communication system based on deep neural networks, the transmitter converts raw data into semantic features and sends the semantic features to the receiver, which receives the semantic features and extracts the semantic information. During the semantic communication process, noise on the transmitter side may introduce semantic noise and cause misunderstandings on the receiver side, such as background noise in the input speech or other users' speech. Adversarial training is a semantic communication method that combats semantic noise: when training the semantic encoder and decoder, noise is added to the input, and the semantic encoder, decoder, and noise are jointly trained. The trained semantic encoder and decoder have the ability to combat noise, but each pair of semantic encoders and decoders in adversarial training needs to be jointly trained for semantic noise, which has poor scalability, and the trained semantic encoder and decoder do not fully utilize the relevant information of semantic noise.

[0120] In view of the above problems, the embodiments of the present application provide a communication method and a communication device, which can improve the impact of noise at the transmitting end on the accuracy of semantic understanding in a semantic communication system.

[0121] The following will describe in detail the method provided by the embodiments of the present application in conjunction with the accompanying drawings. The embodiments provided by the present application can be applied to the communication system shown in Figure 1 above without limitation.

[0122] In the following embodiments, the first device and the second device are taken as examples for illustrative description.

[0123] Among them, the first device can be a terminal device or a component of a terminal device (such as a chip or circuit), or the first device can be a network device or a component of a network device (such as a chip or circuit), or the first device can be an AI node or a component of an AI node (such as a chip or circuit).

[0124] The second device may be a terminal device or a component of a terminal device (such as a chip or circuit), or the second device may be a network device or a component of a network device (such as a chip or circuit), or the second device may be an AI node or a component of an AI node (such as a chip or circuit).

[0125] FIG5 is a flow chart of a semantic communication system training and application method 500 provided in an embodiment of the present application.

[0126] Please refer to Figure 5. The first device may include a basic semantic encoder, which may include a pre-trained encoder. The second device may include a basic semantic decoder, which may include a pre-trained decoder. The basic semantic encoder and the basic semantic decoder may be obtained through joint training.

[0127] Specifically, please continue to refer to Figure 3. The basic semantic encoder and decoder are constructed based on a neural network. The basic semantic encoder and decoder can be jointly trained on a noise-free dataset. The noise-free dataset can include noise-free input data S and semantic information or labels corresponding to the input data S. During the training process, the semantic features output by the encoder are passed to the decoder after channel coding, channel transmission and channel decoding, and serve as the input of the decoder training. As the target output of decoder training, the weight of the decoder neural network is adjusted, and the encoder can adjust the weight of the encoder neural network according to the training situation of the decoder.

[0128] Specifically, please continue to refer to Figure 3. During the information transmission between the basic semantic encoder and the basic semantic decoder, channel coding and channel decoding can also be implemented through a neural network, wherein the training of the channel encoder and channel decoder can be carried out simultaneously with the training of the basic semantic encoder and basic semantic decoder, and the weights of the channel encoder and channel decoder neural networks are adjusted according to the training status of the basic semantic decoder.

[0129] Furthermore, the basic semantic encoder and the basic semantic decoder can be jointly trained by the first device and the second device, or trained by the first device or the second device or other nodes and then distributed to the first device and the second device respectively.

[0130] Continuing with FIG5 , in one possible implementation, method 500 includes:

[0131] S501, the first device obtains an enhanced semantic encoder, which specifically includes: the first device trains a first neural network, wherein the noisy fifth data is the input of the first neural network, the fifth semantic feature is the target output of the first neural network, the fifth semantic feature includes the output of the basic semantic encoder with the fifth data as input, the basic semantic encoder is a pre-trained encoder, and the enhanced semantic encoder includes the trained first neural network.

[0132] Specifically, the first neural network can adopt a trained basic semantic encoder, that is, the first device uses the fifth data and the fifth semantic feature to train the basic semantic encoder to obtain an enhanced semantic encoder; the first neural network can also adopt other types of neural networks that are independent of the basic semantic encoder, such as an untrained neural network model or a neural network model trained for other tasks, and the first device uses the fifth data and the fifth semantic feature to train other types of neural network models to obtain an enhanced semantic encoder.

[0133] Specifically, the fifth data may include data used to train the enhanced semantic encoder, for example, image data, voice data, text data, etc. used to train the enhanced semantic encoder. In the process of obtaining the enhanced semantic encoder, the noisy fifth data is used as the training input of the first neural network, or in other words, noise and the fifth data are input to the first neural network, and the target output of the first neural network is the fifth semantic feature.

[0134] Specifically, the fifth semantic feature may include the target output of the first neural network in the process of obtaining the enhanced semantic encoder, the semantic feature may be a vector containing multiple element values, and the fifth semantic feature may include the target value of the feature vector extracted by the first neural network from the noisy fifth data in the process of obtaining the enhanced semantic encoder.

[0135] FIG6 is a schematic diagram of an enhanced semantic encoder training provided in an embodiment of the present application.

[0136] Please refer to Figure 6. In the process of the first device acquiring the enhanced semantic encoder, the input of the enhanced semantic encoder or the first neural network is a noisy signal. The output is the enhanced semantic feature; the input of the basic semantic encoder is the noise-free signal S, S and Correlation, that is, S is a signal without noise The output of the basic semantic encoder is the basic semantic feature; the target output of the first neural network training is the first neural network input The semantic features of the output are similar to the semantic features of the output after the basic semantic encoder input S, or the first neural network input The distance between the semantic features output later and the semantic features output after the basic semantic encoder input S is smaller.

[0137] Please continue to refer to FIG. 6 . As shown in FIG. 6 , the first device may obtain an enhanced semantic encoder for its local data, namely, the noisy data, noise-free data, and basic semantic encoder of the first device.

[0138] In an embodiment of the present application, the enhanced semantic encoder can be obtained using input data and a basic semantic encoder, that is, the acquisition of the enhanced semantic encoder can be completed independently by the encoder end. Compared with the semantic communication system that overcomes noise through joint training, the independently trained encoder can stagger the time of encoder training and decoder training, and flexibly select the time to transmit the data required for training the decoder. In addition, the amount of data used for encoder training and decoder training can also be different, which increases the flexibility of semantic communication system training.

[0139] Furthermore, the enhanced semantic encoder may simultaneously output calibration information, where the calibration information may include a signal-to-noise ratio of the original data or semantic features. Specifically, the calibration information may include a noise level or a distortion position of each semantic feature or input data.

[0140] Specifically, the calibration information can be directly output by the enhanced semantic encoder, or the calibration information can be obtained by processing the input data or output semantic features of the encoder based on another neural network, which can be trained based on the performance of the encoder or decoder, or the difference between the received semantic features and the sent semantic features.

[0141] Exemplarily, the calibration information may include the semantic feature or the noise level of the input data. For example, the calibration information is a decimal between 0 and 1. The closer the calibration information is to 0, the lower the noise level of the semantic feature or the input data. The closer the calibration information is to 1, the higher the noise level of the semantic feature or the input data.

[0142] Exemplarily, the calibration information may include the distorted position of the semantic feature or the input data. For example, the calibration information includes the element number of the semantic feature. For example, the calibration information is {1, 6}, which indicates that the first and sixth elements in the semantic feature are distorted. For example, the calibration information includes a bitmap of the semantic feature distortion, where 1 indicates distortion and 0 indicates no distortion. For example, the calibration information is {1, 0, 0, 1}, which indicates that the first and fourth elements in the semantic feature are distorted.

[0143] It should be understood that the calibration information can also be obtained through the basic semantic encoder without enhancement. For example, the calibration information can be based on the basic semantic features output by the basic semantic encoder, or determined based on the input data. The embodiment of the present application does not limit the type of semantic encoder for obtaining the calibration information.

[0144] Continuing with FIG5 , in one possible implementation, method 500 further includes:

[0145] S502. The first device obtains a second semantic feature and second calibration information through an enhanced semantic encoder, where the second semantic feature includes an output obtained by the enhanced semantic encoder with second data as input, the second calibration information includes the second semantic feature or a signal-to-noise ratio of the second data, and the second data includes noisy data.

[0146] Specifically, the second data can be used to input an enhanced semantic encoder to obtain a second semantic feature. For example, the second data may include image data, voice data, text data, etc. The second semantic feature is used to adjust the basic semantic decoder to obtain an enhanced semantic decoder. In the process of obtaining the second semantic feature, the noisy second data is used as the input of the enhanced semantic encoder, and the enhanced semantic encoder outputs the second semantic feature.

[0147] Specifically, the second semantic feature may include the output obtained after the enhanced semantic encoder inputs the noisy second data, the semantic feature may be a vector containing multiple element values, and the second semantic feature may include a feature vector extracted by the enhanced semantic encoder from the noisy second data.

[0148] Specifically, the second calibration information may include the output obtained after the enhanced semantic encoder inputs the noisy second data, or the second calibration information may include the output obtained by processing the input data or output semantic features of the encoder based on another neural network. The second calibration information can be used together with the second semantic features as the input of the basic semantic decoder in the second device, and is used to adjust the basic semantic decoder to obtain an enhanced semantic decoder.

[0149] Continuing with FIG5 , in one possible implementation, method 500 further includes:

[0150] S503, the first device sends the second semantic feature, the second calibration information and the first semantic information, and accordingly, the second device receives the second semantic feature, the second calibration information and the first semantic information, the first semantic information includes a label corresponding to the second data, and the second semantic feature, the second calibration information and the first semantic information are used to adjust the basic semantic decoder to obtain an enhanced semantic decoder.

[0151] Specifically, the first semantic information is associated with the second data input to the enhanced semantic encoder, the first semantic information may include semantic information corresponding to the second data, and the first semantic information may be used as the target output of the semantic decoder training to adjust the basic semantic decoder to obtain the enhanced semantic decoder.

[0152] Exemplarily, the second data includes a piece of text or voice data containing language, such as a comment posted by a user on the Internet, and the first semantic information includes the attitude expressed by the text or voice data in the second data, such as approval, criticism, positive, negative, active, passive, etc.

[0153] Exemplarily, the second data includes a batch of image data, such as photos containing human faces, and the first semantic information includes the attitude expressed by the text or voice data corresponding to the second data, such as the gender, age, expression, etc. of the person in the photo.

[0154] It should be understood that the examples of the second data and the first semantic information are only used to help understand the technical solutions of the embodiments of the present application, and are not intended to limit the embodiments of the present application. The content regarding data and semantic information in the following text can refer to the above examples.

[0155] In an embodiment of the present application, the enhanced semantic encoder can be an encoder independently trained with noisy data. The first device sends the first semantic information and the second semantic features and second calibration information obtained by the enhanced semantic encoder using the second data as input, so that the second device that receives the second semantic features, the second calibration information and the first semantic information can use the calibration information to assist the second semantic features and the first semantic information to adjust the basic semantic decoder to obtain an enhanced semantic decoder, so that the semantic communication system can use the second calibration information to train a more reliable decoder.

[0156] Continuing with FIG5 , in one possible implementation, method 500 further includes:

[0157] S504: The second device uses the second calibration information to assist the second semantic features and the first semantic information in adjusting the basic semantic decoder to obtain an enhanced semantic decoder. This includes training the basic semantic decoder, wherein the second semantic features and the second calibration information serve as inputs to the basic semantic decoder, the first semantic information is a target output of the basic semantic decoder, and the second semantic features, the second calibration information, and the first semantic information are used to adjust weights of a neural network of the basic semantic decoder.

[0158] FIG7 is a decoding diagram of an enhanced semantic decoder provided in an embodiment of the present application.

[0159] Please refer to Figure 7, the enhanced semantic features and calibration information can be used as input to enhance the semantic decoder. It can be used as the target output for enhancing semantic decoder training and adjusting the weights of the decoder neural network.

[0160] Specifically, the second calibration information can be cascaded with the second semantic feature as a training input for adjusting the basic semantic decoder to obtain an enhanced semantic decoder, or as an input for the enhanced semantic decoder during semantic reasoning, and the first semantic information can be used as the target output of decoder training to adjust the weights of the decoder neural network.

[0161] Specifically, the second calibration information can be decoded by a second neural network assisted decoder, the input of the second neural network is the second calibration information, the output of the second neural network is associated with some layers of the decoder neural network, and the second neural network is used to adjust the weights of some layers of the decoder through the calibration information.

[0162] Specifically, the second calibration information can be decoded by a third neural network assisted decoder, the input of the third neural network is the second semantic feature and the second calibration information, the output of the third neural network is the input of the decoder neural network, and the third neural network is used to preprocess the second semantic feature based on the second calibration information.

[0163] Specifically, the second calibration information is processed by a neural network, and in the process of assisting the enhanced semantic decoder in decoding, it can refer to fine-tuning some neural network parameters of the decoder, for example, the output of the second neural network or the third neural network is connected to some layers of the decoding neural network to adjust the weights of some layers.

[0164] In an embodiment of the present application, the decoder can use the second semantic feature, the second calibration information and the first semantic information to adjust the weight of the basic semantic decoder during decoding, and the second calibration information can be used to increase the reliability of the adjusted decoder.

[0165] Continuing with FIG5 , in one possible implementation, method 500 further includes:

[0166] S505: The first device obtains a first semantic feature and first calibration information through an enhanced semantic encoder, where the first semantic feature includes an output of the enhanced semantic encoder using the first data as input, the first calibration information includes a signal-to-noise ratio of the first semantic feature or the first data, the first data includes noisy data, and the enhanced semantic encoder is a semantic encoder trained using the noisy data.

[0167] Specifically, the first semantic feature may include the semantic feature output by the enhanced semantic encoder during the semantic reasoning process. The semantic feature may be a vector containing multiple element values. The first semantic feature may include the feature vector extracted by the enhanced semantic encoder from the noisy first data during the semantic reasoning process.

[0168] Specifically, the first data may include data of semantic information to be acquired collected by the first device, such as image data, voice data, text data, etc. of the semantic information to be acquired. In the process of acquiring semantic information, the noisy first data is used as input of the enhanced semantic encoder, or in other words, noise and the first data are input to the enhanced semantic encoder, the enhanced semantic encoder outputs the first semantic feature, and the first device obtains the first calibration information.

[0169] Specifically, the first calibration information may include a signal-to-noise ratio of the first data or the first semantic feature. For example, the first calibration information may include a noise level or a distortion position of the first semantic feature or the first data.

[0170] Specifically, the first calibration information can be directly output by the enhanced semantic encoder, or the first calibration information can also be obtained by processing the first data or the first semantic feature based on another neural network, which can be trained based on the performance of the enhanced semantic encoder or the enhanced semantic decoder, or the difference between the received semantic feature and the sent semantic feature.

[0171] Exemplarily, the calibration information may include the noise level of the first semantic feature or the first data. For example, the calibration information is a decimal between 0 and 1. The closer the calibration information is to 0, the lower the noise level of the first semantic feature or the first data. The closer the calibration information is to 1, the higher the noise level of the first semantic feature or the first data.

[0172] Exemplarily, the calibration information may include the distorted position of the first semantic feature or the first data. For example, the first calibration information includes the element sequence number of the first semantic feature. For example, the calibration information is {1, 6}, indicating that the first and sixth elements in the first semantic feature are distorted. For example, the calibration information includes a bitmap of the semantic feature distortion, where 1 indicates distortion and 0 indicates no distortion. For example, the calibration information is {1, 0, 0, 1}, indicating that the first and fourth elements in the semantic feature are distorted.

[0173] Continuing with FIG5 , in one possible implementation, method 500 further includes:

[0174] S506: The first device sends the first semantic feature and the first calibration information, and correspondingly, the second device receives the first semantic feature and the first calibration information.

[0175] S507: The second device decodes semantic information corresponding to the first data using the first semantic feature and the first calibration information.

[0176] Specifically, the first calibration information can be cascaded with the first semantic feature and used as input to enhance the semantic decoder during semantic reasoning, thereby assisting the decoder in acquiring semantic information more accurately.

[0177] Specifically, the first calibration information can be decoded by a second neural network to assist the decoder, the input of the second neural network is the first calibration information, the output of the second neural network is associated with a partial layer of the decoder neural network, and the second neural network is used to send information that is beneficial to auxiliary decoding to a partial layer of the decoder through the first calibration information.

[0178] Specifically, the first calibration information can be decoded by a third neural network assisted decoder, the input of the third neural network is the first semantic feature and the first calibration information, the output of the third neural network is the input of the decoder neural network, and the third neural network is used to preprocess the first semantic feature based on the first calibration information.

[0179] In an embodiment of the present application, a first device sends a first semantic feature and first calibration information, wherein the first semantic feature and the first calibration information may be information obtained after the enhanced semantic encoder in the first device inputs noisy data, and the enhanced semantic encoder may be an encoder independently trained with noisy data. The second device that receives the first semantic feature and the first calibration information may use the first calibration information to assist the decoder in decoding the first semantic feature, thereby increasing the reliability of the decoder in the second device, so that the semantic communication system can more accurately obtain the semantic information corresponding to the first data when the input first data contains noise.

[0180] FIG8 is a semantic communication system module provided in an embodiment of the present application.

[0181] Please refer to Figure 8. The upper part of Figure 8 includes the system modules of the basic semantic encoder and the basic semantic decoder. For its specific structure, please refer to the description related to Figure 3 above. The lower part of Figure 8 includes the system modules of the enhanced semantic encoder and the enhanced semantic decoder, wherein the enhanced semantic encoder can be obtained from the basic semantic encoder, the enhanced semantic decoder can be obtained from the basic semantic decoder, and the enhanced semantic encoder inputs noisy data. After that, the corresponding semantic features and calibration information can be obtained. The semantic features and calibration information can be transmitted to the enhanced semantic decoder through the transmission module respectively, and the semantic information corresponding to the noisy data can be obtained as the input of the enhanced semantic decoder.

[0182] In a possible implementation, the first calibration information and / or the second calibration information are processed using a cyclic redundancy check (CRC) and / or a retransmission mechanism when being sent.

[0183] Specifically, the transmission modes of the semantic features and calibration information transmitted to the enhanced semantic decoder through the transmission module can be different. The semantic features can be transmitted based on the channel coding and decoding of the neural network, and the calibration information can be transmitted based on the traditional channel coding and decoding.

[0184] Exemplarily, the transmission module of semantic features directly maps the semantic features into symbols, and sends them to the decoder after waveform shaping; the transmission module of calibration information represents the calibration information as bits, and sends it after one or more processes including CRC encoding, channel coding, modulation, and waveform shaping. CRC error checking and retransmission mechanism can be used during the processing process.

[0185] In an embodiment of the present application, the calibration information can be transmitted from the first device to the second device in a more reliable mode by adopting a cyclic redundancy check CRC and / or a retransmission mechanism, thereby ensuring the accuracy of the calibration information when used for auxiliary decoding and / or for auxiliary training of the decoder, and improving the performance of reasoning and training of the semantic communication system.

[0186] FIG9 is a flow chart of another semantic communication system training and application method 900 provided in an embodiment of the present application.

[0187] Please refer to Figure 9. In Figure 9, the first device includes a basic semantic encoder, and the second device includes a basic semantic decoder. The training and construction of the basic semantic encoder and the basic semantic decoder have been described in detail above, and the embodiments of the present application will not be repeated here.

[0188] Continuing with FIG9 , in one possible implementation, method 900 includes:

[0189] S901, determine whether to obtain an enhanced semantic encoder: when the distance between the sixth semantic feature and the seventh semantic feature is greater than or equal to a first threshold, obtain an enhanced semantic encoder; the sixth semantic feature includes the output of the basic semantic encoder with the sixth data as input, the seventh semantic feature includes the output of the basic semantic encoder with the noisy sixth data as input, and the basic semantic encoder is a pre-trained encoder.

[0190] Specifically, the sixth data may include test data used to determine whether to enhance the semantic encoder, for example, image data, voice data, text data, etc. used to determine whether to enhance the semantic encoder.

[0191] Specifically, the sixth semantic feature and the seventh semantic feature may respectively include outputs of a basic semantic encoder when the sixth data and the noisy sixth data are input.

[0192] Specifically, the distance between the sixth semantic feature and the seventh semantic feature may include cosine distance or Euclidean distance, etc. When the distance between the sixth semantic feature and the seventh semantic feature is greater than or equal to the first threshold, it may indicate that the basic semantic encoder is greatly affected by noise, and the first device obtains an enhanced semantic encoder. Optionally, the first device sends a second indication information to the second device, and the second indication information is used to indicate that the encoder of the second device is an enhanced semantic encoder; when the distance between the sixth semantic feature and the seventh semantic feature is less than the first threshold, it may indicate that the basic semantic encoder is less affected by noise, and the first device does not need to obtain an enhanced semantic encoder. Optionally, the first device sends a second indication information to the second device, and the second indication information is used to indicate that the encoder of the second device is a basic semantic encoder.

[0193] In an embodiment of the present application, whether the first device obtains the enhanced semantic encoder is determined based on the distance between the semantic features output when the basic semantic encoder in the first device inputs noisy data and noise-free data. That is, the first device can reasonably judge whether it is necessary to train the enhanced semantic encoder based on the performance of the basic semantic encoder in processing noise, thereby increasing the flexibility of training the semantic communication system.

[0194] S902. Optionally, when the distance between the sixth semantic feature and the seventh semantic feature is greater than or equal to a first threshold, the first device obtains an enhanced semantic encoder. The method for obtaining the enhanced semantic encoder has been described in detail in the above S501 section and will not be repeated in this embodiment of the present application.

[0195] Continuing with FIG9 , in one possible implementation, method 900 further includes:

[0196] S903, optionally, the first device sends a third semantic feature and a fourth semantic feature, and accordingly, the second device receives the third semantic feature and the fourth semantic feature, the third semantic feature and the fourth semantic feature are used to determine the first indication information, the third semantic feature includes the output of the enhanced semantic encoder with the third data as input, and the fourth semantic feature includes the output of the enhanced semantic encoder with the noisy third data as input.

[0197] Specifically, the first indication information may indicate whether the decoder needs to be adjusted, and the first indication information may be used to enable the first device to determine whether the first device sends the second semantic feature and the second calibration information to the second device.

[0198] Specifically, the first indication information may indicate the type of data required for the decoder to adjust, and the first indication information may be used to enable the first device to adjust the second semantic feature and specific content included in the second calibration information.

[0199] Specifically, the third data may include test data used to determine the first indication information, for example, image data, voice data, text data, etc. used to determine the first indication information.

[0200] Specifically, the third semantic feature and the fourth semantic feature may respectively include outputs of an enhanced semantic encoder or a basic semantic encoder when the third data and the noisy third data are input.

[0201] S903, optionally, the first device sends a fourth semantic feature and second semantic information, and accordingly, the second device receives the fourth semantic feature and the second semantic information, the fourth semantic feature and the second semantic information are used to determine the first indication information, the fourth semantic feature includes the output of the enhanced semantic encoder with the noisy third data as input, and the second semantic information includes a label corresponding to the third data.

[0202] It should be understood that when the first device does not enhance the basic semantic encoder, the third semantic feature may include the output of the basic semantic encoder with the third data as input, and the fourth semantic feature includes the output of the basic semantic encoder with the noisy third data as input.

[0203] Specifically, the second semantic information is associated with the third data input to the enhanced semantic encoder or the basic semantic encoder. The second semantic information may include semantic information corresponding to the third data. The second semantic information may be used as the target output for semantic decoder training to adjust the basic semantic decoder to obtain an enhanced semantic decoder.

[0204] Continuing with FIG9 , in one possible implementation, method 900 further includes:

[0205] S904, optionally, the first device sends multiple calibration information types, and correspondingly, the second device receives multiple calibration information types, the first indication information is also used to indicate that the second calibration information is the first type of calibration information among the multiple calibration information types, the second semantic feature includes a semantic feature associated with the first type of calibration information, and the first semantic information includes semantic information associated with the first type of calibration information.

[0206] FIG10 is a schematic diagram of obtaining a calibration information type provided in an embodiment of the present application.

[0207] Please refer to Figure 10. In the process of obtaining multiple calibration information types, the first device can input the noise-free data and the noisy data into the enhanced semantic encoder respectively to obtain noise-free enhanced semantic features and noisy enhanced semantic features respectively, wherein the noise-free data and the noisy data can include multiple batches of data, the noise-free enhanced semantic features and the noisy enhanced semantic features can include multiple strings of semantic feature vectors, and a batch of data can correspond to one or more strings of feature vectors.

[0208] Specifically, when the distance between the noisy enhanced semantic features and the noiseless enhanced semantic feature elements of a batch of data is evenly distributed, it can be judged that the semantic noise in the batch of data or the noisy enhanced semantic features corresponding to the batch of data is evenly distributed, and the calibration information can be used to indicate the noise intensity of each element or the average noise intensity in the noisy enhanced semantic features. The calibration information obtained through the batch of data corresponds to a type with higher accuracy. For example, the calibration information obtained through the batch of data corresponds to a type close to type 1 in Figure 10.

[0209] Specifically, when the distance distribution between the noisy enhanced semantic features and the noiseless enhanced semantic feature elements of a batch of data is uneven, for example, the distance distribution is that the distance between some noisy enhanced semantic features and the noiseless enhanced semantic feature elements is large, and the distance between some noisy enhanced semantic features and the noiseless enhanced semantic feature elements is small, it can be judged as burst semantic noise, and the calibration information can be used to indicate the semantic feature 0 / 1 mask and noise intensity, for example, assigning a weight of 1 to the semantic feature with large noise and a weight of 0 to the semantic feature with small noise. The calibration information obtained through this batch of data corresponds to a type with lower accuracy. For example, the calibration information obtained through this batch of data corresponds to a type close to type K in Figure 10.

[0210] Exemplarily, the input of the enhanced semantic encoder includes 5 batches of data. The first device corresponds the calibration information obtained by the enhanced semantic encoder for the 5 batches of data to 3 calibration information types based on the noise distribution of the 5 batches of data, wherein the {0, 1}th batch of data corresponds to type 1, and the noise in the {0, 1}th batch of data is uniformly distributed; the {3, 4}th batch of data corresponds to type 3, and the {3, 4}th batch of data is burst noise; the {2}th batch of data corresponds to type 2, and the noise distribution in the {2}th batch of data is between the {0, 1}th batch of data and the {3, 4}th batch of data. The first device sends the above 3 calibration information types to the second device.

[0211] It should be understood that S903 and S904 can occur at the same time, or be sent from the first device to the second device through the same signaling, that is, the first device simultaneously sends the third semantic feature, the fourth semantic feature and multiple calibration information types to the second device, or the first device simultaneously sends the fourth semantic feature, the second semantic information and multiple calibration information types to the second device.

[0212] It should be understood that when the first device does not enhance the basic semantic encoder, the second calibration information, the first type of calibration information, and multiple calibration information types can be based on the basic semantic features output by the basic semantic encoder, or determined based on the input data; when it is determined in S901 that there is no need to enhance the basic semantic encoder, the embodiment of obtaining the second calibration information, the first type of calibration information, and multiple calibration information types through enhanced semantic features in S904 can be replaced by implementation through basic semantic features or through input data.

[0213] Continuing with FIG9 , in one possible implementation, method 900 further includes:

[0214] S905, optionally, determining whether to obtain an enhanced semantic decoder: when a first distance between the third semantic information and the fourth semantic information is greater than or equal to a second threshold, determining that the basic semantic decoder needs to be adjusted, the third semantic information includes the output of the basic semantic decoder with the third semantic feature as input, the fourth semantic information includes the output of the basic semantic decoder with the fourth semantic feature as input, and the first distance includes the cosine distance or Euclidean distance between the third semantic information and the fourth semantic information.

[0215] S905, optionally, determining whether to obtain an enhanced semantic decoder: when the first distance between the second semantic information and the fourth semantic information is greater than or equal to a second threshold, determining that the basic semantic decoder needs to be adjusted, the fourth semantic information includes the output of the basic semantic decoder with the fourth semantic feature as input, and the first distance includes the cosine distance or Euclidean distance between the second semantic information and the fourth semantic information.

[0216] Specifically, please combine S903, the second semantic information includes a label corresponding to the third data, the third semantic information includes semantic information obtained by the noise-free third data through the semantic codec, which can be approximated as the label corresponding to the third data, and the fourth semantic information is the output of the noisy third data through the encoder and the basic semantic decoder. The second device can determine whether to send the first indication information based on the first distance between the fourth semantic information and the second or third semantic information, and request to adjust the basic semantic decoder through the first indication information to obtain the second semantic features and second calibration information required by the enhanced semantic decoder.

[0217] Specifically, the first distance may include the Euclidean distance or cosine distance between semantic information. For example, the semantic information is converted into a vector through an algorithm, and the Euclidean distance or cosine distance between the vectors is calculated.

[0218] In an embodiment of the present application, whether the decoder needs to be adjusted can be judged based on the distance between the second semantic information and the fourth semantic information or the distance between the third semantic information and the fourth semantic information. That is, the decoder can flexibly choose whether to make adjustments based on its own performance in processing noise, thereby improving the flexibility of semantic communication system training.

[0219] In a possible implementation manner, the first type is determined according to the first distance.

[0220] FIG11 is a schematic diagram of calibration information type selection provided in an embodiment of the present application.

[0221] Please refer to Figure 11. The second device calculates the distance between the semantic information decoded from the noisy data and the semantic information or semantic label decoded from the noise-free data, and maps the distance value to one of multiple distance ranges. The multiple distance ranges correspond to multiple optional calibration information types received by the second device. The second device determines the first type based on the distance range mapped to the distance value. The second device can request the first type of calibration information and the semantic features and labels associated with the first type of calibration information from the first device through the first indication information to adjust the basic semantic decoder to obtain an enhanced semantic decoder.

[0222] In an embodiment of the present application, a first device sends multiple calibration information types to a second device. The second device can specifically select the first type from the multiple calibration information types based on its own performance, and inform the first device of the first type through a first indication message. Then, the first device can only send the first type of calibration information, as well as semantic features and semantic information tags associated with the first type of calibration information, to the second device for adjustment of the decoder in the second device. It can be seen that the second device can select some data with better effects to adjust the decoder based on its own performance, thereby reducing the time and information transmission resources required for decoder adjustment and increasing the adjustment efficiency of the decoder.

[0223] Continuing with FIG9 , in one possible implementation, method 900 further includes:

[0224] S906, optionally, the first device receives first indication information, and accordingly, the second device sends the first indication information, where the first indication information is used to request the second semantic feature, the second calibration information, and the first semantic information.

[0225] Specifically, when the distance between the second semantic information and the fourth semantic information or the distance between the third semantic information and the fourth semantic information is less than the second threshold, the second device may not send the first indication information, or the second device may indicate to the first device through the first indication information that it does not need to send training data.

[0226] Specifically, the first indication information may include the calibration information type determined by the second device, and is used to indicate that the training data subsequently sent by the first device is training data corresponding to the first type.

[0227] In an embodiment of the present application, the first device specifically determines whether to send the second semantic feature, the second calibration information and the first semantic information based on the first indication information sent by the second device. Since the second semantic feature, the second calibration information and the first semantic information are used to adjust the basic semantic decoder to obtain an enhanced semantic decoder, the decoder in the second device can determine the first indication information based on its own performance, and inform the first device decoder whether it needs to be adjusted through the first indication information. If the decoder needs to be adjusted, the first device including the encoder will send the second semantic feature, the second calibration information and the first semantic information to the second device, so that the adjustment of the decoder in the semantic communication system is more flexible.

[0228] S907. Optionally, when the first indication information indicates that the basic semantic decoder needs to be adjusted, the first device obtains the second semantic feature and the second calibration information. The method of S907 has been explained in detail in the previous S502 section and will not be repeated here in the embodiment of the present application.

[0229] S908. Optionally, when the first indication information indicates that the basic semantic decoder needs to be adjusted, the first device sends the second semantic feature, the second calibration information, and the first semantic information.

[0230] Furthermore, when the first indication information indicates the first type, what is sent in S908 is calibration information corresponding to the first type, and semantic features and semantic information associated with the calibration information corresponding to the first type.

[0231] It should be understood that the second calibration information and the calibration information corresponding to the first type can be obtained through an enhanced semantic encoder or a basic semantic encoder without enhancement, based on the result of determining whether to enhance the basic semantic encoder in S901; for example, when the basic semantic encoder is not enhanced, the calibration information can be based on the basic semantic features output by the basic semantic encoder, or determined based on the input data. The embodiment of the present application does not limit the type of semantic encoder for obtaining the calibration information.

[0232] It should be understood that S909 to S912 correspond to the contents of S504 to S507 in the previous text. The relevant embodiments have been described in detail in the previous text, and the embodiments of this application will not be repeated here.

[0233] It should be understood that when the basic semantic encoder is not enhanced in method 900 , the steps or actions implemented by the enhanced semantic encoder in method 500 may be implemented by the basic semantic encoder.

[0234] It should also be understood that when the basic semantic decoder is not adjusted in method 900 , the steps or actions implemented by the enhanced semantic decoder in method 500 may be implemented by the basic semantic decoder.

[0235] It is understood that the above steps are merely examples and are not limiting.

[0236] It can be understood that some optional features in the various embodiments of the present application may not depend on other features in certain scenarios, and may also be combined with other features in certain scenarios, without limitation.

[0237] It is also understood that in some of the above embodiments, sending information is mentioned multiple times. Taking A sending information to B as an example, A sending information to B may include A sending information directly to B or A sending information to B through other devices or network elements, and there is no limitation on this.

[0238] It can also be understood that the solutions in the various embodiments of the present application can be reasonably combined and used, and the explanations or descriptions of the various terms appearing in the embodiments can be referenced or explained with each other in the various embodiments, without limitation to this.

[0239] The method provided in the embodiments of the present application is described in detail above with reference to Figures 5 to 11. Below, the apparatus provided in the embodiments of the present application is described in detail with reference to Figures 12 to 15. It should be understood that the description of the apparatus embodiment corresponds to the description of the method embodiment. Therefore, for matters not described in detail, reference can be made to the method embodiment above, and for the sake of brevity, they will not be repeated here.

[0240] FIG12 is a schematic diagram of a functional module of a semantic communication system chip provided in an embodiment of the present application.

[0241] Referring to Figure 12 , the acquisition module in the first device is used to acquire data to be transmitted, the semantic encoder module is used to process the data to obtain semantic features and calibration information, channel encoders 1 and 2 are used to channel-encode the calibration information and semantic features, and the transceiver unit is used to transmit the encoded semantic features and calibration information to the receiving end. The transceiver unit in the second device is used to receive signals of the semantic features and calibration information, channel decoder 1 and channel encoder 2 are used to channel-decode the semantic features and calibration information, the semantic decoder is used to decode semantic information from the semantic features, and the task module is used to perform downstream tasks based on the semantic information.

[0242] FIG13 is a schematic structural diagram of a communication device provided in an embodiment of the present application.

[0243] The communication device 1300 includes a transceiver unit 1310 and a processing unit 1320 , wherein the transceiver unit 1310 can be used to implement corresponding communication functions, and the processing unit 1320 can be used to perform data processing.

[0244] Optionally, the transceiver unit 1310 may also be referred to as a communication interface or communication unit, and may include a transmitting unit and / or a receiving unit. The transceiver unit 1310 may be a transceiver (including a transmitter and / or a receiver), an input / output interface (including an input and / or output interface), a pin, or a circuit. The transceiver unit 1310 may be configured to perform the transmitting and / or receiving steps in the above-described method embodiments.

[0245] Optionally, the processing unit 1320 may be a processor (may include one or more), a processing circuit with processor functions, etc., and may be used to execute other steps except sending and receiving in the above method embodiment.

[0246] Optionally, the communication device 1300 further includes a storage unit, which may be a memory, an internal storage unit (e.g., a register, a cache, etc.), an external storage unit (e.g., a read-only memory, a random access memory, etc.). The storage unit is used to store instructions, and the processing unit 1320 executes the instructions stored in the storage unit to enable the communication device to perform the above method.

[0247] In one design, the communication device 1300 may be used to perform the actions performed by the first device in each of the above method embodiments, for example, the communication device 1300 may be used to perform the actions performed by the first device in the above method 500 or 900. In this case, the communication device 1300 may be a component of the first device, the transceiver unit 1310 may be used to perform the transceiver-related operations on the first device side in the above method embodiments, and the processing unit 1320 may be used to perform the processing-related operations of the first device in the above method embodiments.

[0248] For example, the transceiver unit 1310 is configured to send the first semantic feature and the first calibration information; the processing unit 1320 is configured to obtain the first semantic feature and the first calibration information through the enhanced semantic encoder.

[0249] It should be understood that the transceiver unit 1310 and the processing unit 1320 may also perform other operations performed by the first device in any of the above methods, which will not be described in detail here.

[0250] In one design, the communication device 1300 can be used to perform the actions performed by the second device in each of the above method embodiments, for example, the communication device 1300 can be used to perform the actions performed by the second device in the above method 500 or 900. In this case, the communication device 1300 can be a component of the second device, the transceiver unit 1310 is used to perform the transceiver-related operations on the second device side in the above method embodiments, and the processing unit 1320 is used to perform the processing-related operations of the second device in the above method embodiments.

[0251] For example, the transceiver unit 1310 is configured to receive a first semantic feature and first calibration information; and the processing unit 1320 is configured to decode semantic information corresponding to the first data according to the first semantic feature and the first calibration information.

[0252] It should be understood that the transceiver unit 1310 and the processing unit 1320 may also perform other operations performed by the second device in any of the above methods 500 or 900, which are not described in detail here.

[0253] It should also be understood that the communication device 1300 here is embodied in the form of a functional unit. The term "unit" here can refer to an application specific integrated circuit (ASIC), an electronic circuit, a processor (such as a shared processor, a dedicated processor or a group processor, etc.) and a memory for executing one or more software or firmware programs, a combined logic circuit and / or other suitable components that support the described functions. In an optional example, those skilled in the art will understand that the communication device 1300 can be specifically the second device in the above-mentioned embodiment, and can be used to execute the various processes and / or steps corresponding to the second device in the above-mentioned method embodiments. To avoid repetition, they are not described here.

[0254] The communication device 1300 of each of the above-mentioned solutions has the function of implementing the corresponding steps performed by the device in the above-mentioned method, or the communication device 1300 of each of the above-mentioned solutions has the function of implementing the corresponding steps performed by the access network device in the above-mentioned method. The function can be implemented by hardware, or the corresponding software can be implemented by hardware. The hardware or software includes one or more modules corresponding to the above-mentioned functions; for example, the transceiver unit can be replaced by a transceiver (for example, the sending unit in the transceiver unit can be replaced by a transmitter, and the receiving unit in the transceiver unit can be replaced by a receiver), and other units, such as the processing unit, can be replaced by a processor to respectively perform the sending and receiving operations and related processing operations in each method embodiment.

[0255] In addition, the transceiver unit 1310 may also be a transceiver circuit (for example, may include a receiving circuit and a sending circuit), and the processing unit may be a processing circuit.

[0256] It should be noted that the apparatus in FIG13 may be a network element or device in the aforementioned embodiment, or may be a chip or chip system, such as a system on chip (SoC). The transceiver unit may be an input / output circuit or a communication interface; the processing unit may be a processor, microprocessor, or integrated circuit integrated on the chip. This is not limited here.

[0257] FIG14 is a schematic diagram of a communication architecture provided in an embodiment of the present application.

[0258] The communication device 1400 shown in Figure 14 includes a processor 1410 and, optionally, one or more of a memory 1420 or a transceiver 1430. The processor 1410 is coupled to the memory 1420 and configured to execute instructions stored in the memory 1420 to control the transceiver 1430 to send and / or receive signals.

[0259] It should be understood that the processor 1410 and memory 1420 described above can be combined into a single processing device, with the processor 1410 configured to execute program code stored in the memory 1420 to implement the aforementioned functions. In a specific implementation, the memory 1420 can also be integrated into the processor 1410 or independent of the processor 1410. It should be understood that the processor 1410 can also correspond to the various processing units in the aforementioned communication device, and the transceiver 1430 can correspond to the various receiving units and transmitting units in the aforementioned communication device.

[0260] It should also be understood that the transceiver 1430 may include a receiver (or receiver) and a transmitter (or transmitter). The transceiver may further include an antenna, and the number of antennas may be one or more. The transceiver may also be a communication interface or interface circuit.

[0261] Specifically, the communication device 1400 may correspond to the first device in the method 500 or 900 according to the embodiment of the present application. The communication device 1400 may execute the steps performed by the first device in the method 500 or 900; the communication device 1400 may correspond to the second device in the method 500 or 900 according to the embodiment of the present application. The communication device 1400 may execute the steps performed by the second device in the method 500 or 900. It should be understood that the specific processes of the above-mentioned corresponding steps have been described in detail in the above-mentioned method embodiments and will not be repeated here for the sake of brevity.

[0262] When the communication device 1400 is a chip, the chip includes an interface unit and a processing unit, wherein the interface unit may be an input / output circuit or a communication interface; and the processing unit may be a processor, microprocessor, or integrated circuit integrated on the chip.

[0263] It should be understood that the processor mentioned in the embodiments of the present application may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0264] It should also be understood that the memory mentioned in the embodiments of the present application may be a volatile memory and / or a non-volatile memory. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM). For example, RAM can be used as an external cache. By way of example and not limitation, RAM includes the following forms: static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0265] It should be noted that when the processor is a general-purpose processor, DSP, ASIC, FPGA or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, the memory (storage module) can be integrated into the processor.

[0266] It should also be noted that the memory described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0267] 15 is a schematic diagram of a chip system 1500 provided in an embodiment of the present application. The chip system 1500 (or also referred to as a processing system) includes a logic circuit 1510 and an input / output interface 1520 .

[0268] Logic circuit 1510 may be a processing circuit within chip system 1500. Logic circuit 1510 may be coupled to a storage unit and invoke instructions within the storage unit, enabling chip system 1500 to implement the methods and functions of various embodiments of the present application. Input / output interface 1520 may be an input / output circuit within chip system 1500, outputting information processed by chip system 1500 or inputting data or signaling information to be processed into chip system 1500 for processing.

[0269] As a solution, the chip system 1500 is used to implement the operations performed by the communication device (such as the first device, and such as the second device) in the above various method embodiments.

[0270] For example, the logic circuit 1510 is used to implement the processing-related operations performed by the communication device (such as the first device, and also the second device) in the above method embodiments; the input / output interface 1520 is used to implement the sending and / or receiving-related operations performed by the communication device (such as the first device, and also the second device) in the above method embodiments.

[0271] An embodiment of the present application further provides a computer-readable storage medium storing computer instructions for implementing the methods executed by a communication device (such as the first device or the second device) in the above-mentioned method embodiments.

[0272] For example, when the computer program is executed by a computer, the computer can implement the method performed by the communication device (such as the first device, and such as the second device) in each embodiment of the above method.

[0273] An embodiment of the present application further provides a computer program product comprising instructions, which, when executed by a computer, implement the methods performed by a communication device (such as the first device or the second device) in the above-mentioned method embodiments.

[0274] The present application also provides a communication system, which includes the first device and / or the second device in each of the above embodiments. For example, the system includes the first device and the second device in Figure 5 or Figure 9.

[0275] The explanation of the relevant contents and beneficial effects of any of the above-mentioned devices can be referred to the corresponding method embodiments provided above, which will not be repeated here.

[0276] It should be noted that the processor in the embodiments of the present application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiment can be completed by an integrated logic circuit of the hardware in the processor or by instructions in the form of software. The above processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0277] The present application also provides a computer-readable medium having a computer program stored thereon, which implements the functions of any of the above method embodiments when executed by a computer.

[0278] The present application also provides a computer program product, which implements the functions of any of the above method embodiments when executed by a computer.

[0279] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a high-density digital video disc (DVD)), or a semiconductor medium (eg, a solid state disk (SSD)).

[0280] In the embodiments of this application, words such as "exemplary" and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described in this application as "exemplary" should not be construed as preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete way.

[0281] It should be understood that references to "embodiments" throughout this specification mean that a particular feature, structure, or characteristic associated with the embodiment is included in at least one embodiment of the present application. Therefore, various embodiments throughout this specification do not necessarily refer to the same embodiment. Furthermore, these particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0282] It should be understood that in the various embodiments of the present application, the size of the sequence number of each process does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application. The names of all nodes and messages in this application are merely names set by this application for the convenience of description. The names in the actual network may be different. It should not be understood that this application limits the names of various nodes and messages. On the contrary, any name with the same or similar function as the node or message used in this application is regarded as the method or equivalent replacement of this application, and is within the scope of protection of this application.

[0283] It should also be understood that in this application, "when", "if" and "if" all mean that the UE or base station will take corresponding measures under certain objective circumstances. It does not limit the time, and does not require the UE or base station to take judgment actions when implementing it, nor does it mean that there are other limitations.

[0284] Additionally, the terms "system" and "network" are often used interchangeably. The term "and / or" is simply used to describe an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0285] As used herein, the term "at least one of" or "at least one of" refers to all or any combination of the listed items. For example, "at least one of A, B, and C" can mean: A alone, B alone, C alone, A and B together, B and C together, and A, B, and C together. As used herein, "at least one" means one or more. "A plurality" means two or more.

[0286] It should be understood that the terms "include", "comprising", "having" and their variations mean "including but not limited to", unless specifically emphasized otherwise.

[0287] It should be understood that in various embodiments of the present application, the first, second, and various numerical numbers are merely distinctions for ease of description and are not intended to limit the scope of the embodiments of the present application.

[0288] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0289] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the description of the corresponding processes and beneficial effects in the aforementioned method embodiments, and will not be repeated here.

[0290] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0291] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0292] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0293] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, part of the technical solution of the present application can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0294] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A communication method, characterized in that: include: obtaining, by an enhanced semantic encoder, a first semantic feature and first calibration information, wherein the first semantic feature comprises an output of the enhanced semantic encoder using first data as input, the first calibration information comprises a signal-to-noise ratio of the first semantic feature or the first data, the first data comprises noisy data, and the enhanced semantic encoder is a semantic encoder trained using the noisy data; The first semantic feature and the first calibration information are sent, where the first semantic feature and the first calibration information are used to decode semantic information corresponding to the first data.

2. The method according to claim 1, characterized in that Before acquiring the first semantic feature and the first calibration information by the enhanced semantic encoder, the method further includes: Obtaining the enhanced semantic encoder; obtaining, by the enhanced semantic encoder, a second semantic feature and second calibration information, wherein the second semantic feature comprises an output of the enhanced semantic encoder with second data as input, the second calibration information comprises the second semantic feature or a signal-to-noise ratio of the second data, and the second data comprises noisy data; The second semantic feature, the second calibration information and the first semantic information are sent, where the first semantic information includes a label corresponding to the second data, and the second semantic feature, the second calibration information and the first semantic information are used to adjust a basic semantic decoder to obtain an enhanced semantic decoder.

3. The method according to claim 2, characterized in that The first calibration information and / or the second calibration information are processed by using a cyclic redundancy check CRC and / or a retransmission mechanism when being sent.

4. The method according to claim 2 or 3, characterized in that Before sending the second semantic feature, the second calibration information, and the first semantic information, the method further includes: First indication information is received, where the first indication information is used to request the second semantic feature, the second calibration information, and the first semantic information.

5. The method according to claim 4, characterized in that Also includes: Send a third semantic feature and a fourth semantic feature, where the third semantic feature and the fourth semantic feature are used to determine the first indication information, the third semantic feature includes the output of the enhanced semantic encoder with the third data as input, and the fourth semantic feature includes the output of the enhanced semantic encoder with the noisy third data as input.

6. The method according to claim 4, characterized in that Also includes: Send a fourth semantic feature and second semantic information, where the fourth semantic feature and the second semantic information are used to determine the first indication information, the fourth semantic feature includes the output of the enhanced semantic encoder with the noisy third data as input, and the second semantic information includes a label corresponding to the third data.

7. The method according to any one of claims 4 to 6, characterized in that Also includes: Multiple calibration information types are sent, and the first indication information is also used to indicate that the second calibration information is the first type of calibration information among the multiple calibration information types, the second semantic feature includes a semantic feature associated with the first type of calibration information, and the first semantic information includes semantic information associated with the first type of calibration information.

8. The method according to any one of claims 2 to 7, characterized in that The obtaining of the enhanced semantic encoder comprises: A first neural network is trained, wherein the noisy fifth data is the input of the first neural network, the fifth semantic feature is the target output of the first neural network, the fifth semantic feature includes the output of a basic semantic encoder with the fifth data as input, the basic semantic encoder is a pre-trained encoder, and the enhanced semantic encoder includes the trained first neural network.

9. The method according to any one of claims 2 to 7, characterized in that When the distance between the sixth semantic feature and the seventh semantic feature is greater than a first threshold, the enhanced semantic encoder is obtained; the sixth semantic feature includes the output of the basic semantic encoder with the sixth data as input, and the seventh semantic feature includes the output of the basic semantic encoder with the noisy sixth data as input, and the basic semantic encoder is a pre-trained encoder.

10. A communication method, characterized in that: include: receiving a first semantic feature and first calibration information, where the first semantic feature and the first calibration information are obtained by an enhanced semantic encoder, the first semantic feature comprising an output of the enhanced semantic encoder using first data as input, the first calibration information comprising a signal-to-noise ratio of the first semantic feature or the first data, the first data comprising noisy data, and the enhanced semantic encoder being a semantic encoder trained using the noisy data; Semantic information corresponding to the first data is decoded using the first semantic feature and the first calibration information.

11. The method according to claim 10, characterized in that Before receiving the first semantic feature and the first calibration information, the method further includes: receiving a second semantic feature, second calibration information, and first semantic information, wherein the second semantic feature comprises an output of the enhanced semantic encoder with second data as input, the second calibration information comprises the second semantic feature or a signal-to-noise ratio of the second data, the second data comprises noisy data, and the first semantic information comprises a label corresponding to the second data; A base semantic decoder is adjusted using the second semantic features, the second calibration information, and the first semantic information to obtain an enhanced semantic decoder.

12. The method according to claim 11, characterized in that The first calibration information and / or the second calibration information are processed by using a cyclic redundancy check CRC and / or a retransmission mechanism when being sent.

13. The method according to claim 11 or 12, characterized in that Before receiving the second semantic feature, the second calibration information, and the first semantic information, the method further includes: First indication information is sent, where the first indication information is used to request the second semantic feature, the second calibration information, and the first semantic information.

14. The method according to claim 13, characterized in that Also includes: A third semantic feature and a fourth semantic feature are received, where the third semantic feature and the fourth semantic feature are used to determine the first indication information, the third semantic feature includes the output of the enhanced semantic encoder with the third data as input, and the fourth semantic feature includes the output of the enhanced semantic encoder with the noisy third data as input.

15. The method according to claim 14, characterized in that When the first distance between the third semantic information and the fourth semantic information is greater than or equal to a second threshold, it is determined that the basic semantic decoder needs to be adjusted, the third semantic information includes the output obtained by the basic semantic decoder with the third semantic feature as input, the fourth semantic information includes the output obtained by the basic semantic decoder with the fourth semantic feature as input, and the first distance includes the cosine distance or Euclidean distance between the third semantic information and the fourth semantic information.

16. The method according to claim 13, characterized in that Also includes: A fourth semantic feature and second semantic information are received, where the fourth semantic feature and the second semantic information are used to determine the first indication information, the fourth semantic feature including the output of the enhanced semantic encoder with noisy third data as input, and the second semantic information including a label corresponding to the third data.

17. The method according to claim 16, characterized in that When the first distance between the second semantic information and the fourth semantic information is greater than or equal to a second threshold, it is determined that the basic semantic decoder needs to be adjusted, the fourth semantic information includes the output obtained by the basic semantic decoder with the fourth semantic feature as input, and the first distance includes the cosine distance or Euclidean distance between the second semantic information and the fourth semantic information.

18. The method according to claim 15 or 17, characterized in that Also includes: Multiple calibration information types are received, and the first indication information is also used to indicate that the second calibration information is the first type of calibration information among the multiple calibration information types, the second semantic feature includes a semantic feature associated with the first type of calibration information, and the first semantic information includes semantic information associated with the first type of calibration information.

19. The method according to claim 18, characterized in that The first type is determined according to the first distance.

20. The method according to any one of claims 11 to 19, characterized in that The adjusting a basic semantic decoder using the second semantic feature, the second calibration information, and the first semantic information to obtain an enhanced semantic decoder includes: Train the basic semantic decoder, wherein the second semantic feature and the second calibration information are inputs of the basic semantic decoder, the first semantic information is the target output of the basic semantic decoder, and the second semantic feature, the second calibration information and the first semantic information are used to adjust the weights of the basic semantic decoder neural network.

21. A communication device, characterized in that: The communication device is configured to execute the method according to any one of claims 1 to 9 or 10 to 20.

22. A communication device, characterized in that: The communication device includes at least one processor, and the at least one processor is configured to execute a computer program or instruction so that the method according to any one of claims 1 to 9 is executed, or the method according to any one of claims 10 to 20 is executed.

23. The communication device according to claim 22, wherein: The communication device further comprises at least one memory for storing the computer program or instructions.

24. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program or instruction. When the computer program or instruction is executed on a computer, the method according to any one of claims 1 to 9 is executed, or the method according to any one of claims 10 to 20 is executed.

25. A computer program product, characterized in that When the computer program product is run on a computer, the method according to any one of claims 1 to 9 is executed, or the method according to any one of claims 10 to 20 is executed.

Images