Communication method and apparatus

Abstract

A communication method and apparatus. The method comprises: receiving a first data packet, wherein data carried by the first data packet can be decoded by M neural network layers, and M is a positive integer; and sending a second data packet and first indication information, wherein the second data packet is obtained by decoding, by means of K neural network layers, the data carried by the first data packet, data carried by the second data packet can be decoded by M-K neural network layers, K is a positive integer, and K<M. The K neural network layers belong to the M neural network layers. The first indication information indicates that decoding has been performed by means of the K neural network layers, or that decoding has not been performed by means of the M-K neural network layers. By using the above method, decoding can be performed by a first device first by means of some neural network layers. Since the first device has more powerful neural network computing capabilities, the decoding efficiency can be improved, and the end-to-end delay of a data packet can be ensured.

Description

Communication method and device

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office of the People's Republic of China on November 22, 2023, with application number 202311574838.7 and invention name "A Communication Method and Device", the entire contents of which are incorporated by reference into this application. Technical Field

[0003] The present application relates to the field of communication technology, and in particular to a communication method and device. Background Art

[0004] Codecs based on artificial intelligence (AI) feature streams can reduce the transmission rate of streaming media communications, particularly for extended reality (XR) services. Current AI feature stream codecs used in industry research have large parameter counts, typically exceeding millions or even tens of millions.

[0005] On the one hand, terminal devices are limited by size and usually cannot deploy powerful computing units, so it is difficult to complete large-scale neural network calculations. On the other hand, XR services have strict end-to-end latency requirements. Generally speaking, from the time the server starts rendering the XR video frame to the time the terminal device displays the corresponding image of the XR video frame, the latency needs to be controlled within 70ms, of which the time it takes to decode the terminal device is usually around 10 to 20ms. Therefore, the computing power level of the terminal device is also difficult to complete the encoding of the AI ​​feature stream within the latency requirements, resulting in a poor user experience.

[0006] Summary of the Invention

[0007] The embodiments of the present application provide a communication method and apparatus to solve the problem that the computing power of a terminal device is limited and cannot efficiently complete the encoding and decoding of an AI feature stream, thereby resulting in a poor user experience.

[0008] In a first aspect, the present application provides a communication method, which can be executed by a first device or a module (such as a chip) in the first device. For example, the first device can be an access network device or a core network element. The method includes: receiving a first data packet; wherein the data carried by the first data packet can be decoded by M neural network layers, M is a positive integer; sending a second data packet and a first indication information, wherein the second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, and the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M, and the K neural network layers belong to the M neural network layers; the first indication information indicates that it has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers.

[0009] Using the above method, the first device can first perform decoding of part of the neural network layer, which can reduce the dependence on the neural network computing capability of the terminal device and save the power consumption and cost of the terminal device. Compared with the terminal device, the first device has stronger neural network computing capability, which can improve the decoding efficiency, reduce the transmission delay of the data packet, and improve the user service experience.

[0010] Among them, the data carried by the first data packet can be decoded by M neural network layers, which can also be understood as, the data carried by the first data packet can be decoded by M neural network layers, or the data carried by the first data packet can be decoded using M neural network layers, or the data carried by the first data packet needs to be decoded using M neural network layers.

[0011] The data carried by the second data packet can be decoded by MK neural network layers. It can also be understood that the data carried by the second data packet can be decoded by MK neural network layers, or the data carried by the second data packet can be decoded using MK neural network layers, or the data carried by the second data packet needs to be decoded using MK neural network layers.

[0012] In one possible design, the first data packet includes second indication information, where the second indication information indicates that the data carried by the first data packet can be decoded by a neural network. Alternatively, the second indication information can be understood as indicating that the data carried by the first data packet can be decoded by a neural network. This design can be used to indicate the encoding and decoding method of the first data packet.

[0013] In one possible design, third indication information is received, where the third indication information indicates that the data carried by the first data packet can be decoded by a neural network. This can also be understood as indicating that the data carried by the first data packet can be decoded by a neural network. This design can be used to indicate the encoding and decoding method of the first data packet.

[0014] In one possible design, fourth indication information is received, and the fourth indication information indicates a mapping relationship between the M neural network layers and the N decoding subtasks, wherein each of the N decoding subtasks corresponds to at least one of the M neural network layers, and N is a positive integer; the second data packet is obtained by executing S of the N decoding subtasks on the data carried by the first data packet, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, and S<N. With the above design, it is possible to execute some decoding subtasks.

[0015] In one possible design, the first indication information indicates that the S decoding subtasks have been completed, or that NS decoding subtasks are uncompleted, the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers, and the NS decoding subtasks are the decoding subtasks of the N decoding subtasks excluding the S decoding subtasks. The above design can be used to indicate completed decoding subtasks or uncompleted decoding subtasks.

[0016] In one possible design, the K neural network layers are determined based on capability information and / or channel state information of the terminal device, wherein the capability information of the terminal device is used to indicate the neural network computing capability of the terminal device.

[0017] It can also be understood that before decoding the data carried by the first data packet through the K neural network layers, the terminal device's capability information and / or channel state information is obtained, where the terminal device's capability information is used to indicate the terminal device's neural network computing capability; and the K neural network layers are determined based on the terminal device's capability information and / or channel state information. The above design can be used to determine which neural network layers need to be decoded.

[0018] In a second aspect, the present application provides a communication method, which can be executed by a terminal device or a module (such as a chip) in the terminal device. The method includes: receiving a second data packet and a first indication information, wherein the data carried by the second data packet can be decoded by MK neural network layers, the K neural network layers belong to M neural network layers, M and K are positive integers, and K<M; the first indication information indicates that the data has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers; decoding the data carried by the second data packet according to the first indication information to obtain service data.

[0019] Using the above method, the terminal device performs decoding of the remaining neural network layers, thereby reducing the dependence on the neural network computing capabilities of the terminal device, saving the power consumption and cost of the terminal device, and improving the user service experience.

[0020] In one possible design, fourth indication information is received, where the fourth indication information indicates a mapping relationship between M neural network layers and N decoding subtasks, where each of the N decoding subtasks corresponds to at least one of the M neural network layers, and N is a positive integer. Using the above design, the terminal device can obtain the mapping relationship between the M neural network layers and the N decoding subtasks.

[0021] In one possible design, the first indication information indicates that S decoding subtasks have been completed, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, S<N, or the first indication information indicates that NS decoding subtasks are not completed, and the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers. The above design enables the terminal device to obtain information about completed decoding subtasks or unfinished decoding subtasks.

[0022] In a third aspect, the present application provides a communication device, comprising: a processing unit and a transceiver unit; the transceiver unit is used to send and receive information; the processing unit is used to receive a first data packet through the transceiver unit; wherein the data carried by the first data packet can be decoded by M neural network layers, M is a positive integer; and the second data packet and first indication information are sent through the transceiver unit, wherein the second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, and the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M; the K neural network layers belong to the M neural network layers; the first indication information indicates that it has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers.

[0023] In one possible design, the first data packet includes second indication information, where the second indication information indicates that data carried by the first data packet can be decoded by a neural network.

[0024] In one possible design, the transceiver unit is used to receive third indication information, where the third indication information indicates that the data carried by the first data packet can be decoded by a neural network.

[0025] In one possible design, the transceiver unit is used to receive fourth indication information, where the fourth indication information indicates a mapping relationship between the M neural network layers and the N decoding subtasks, wherein each of the N decoding subtasks corresponds to at least one of the M neural network layers, and N is a positive integer; the second data packet is obtained by executing S of the N decoding subtasks on the data carried by the first data packet, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, and S<N.

[0026] In one possible design, the first indication information indicates that the S decoding subtasks have been completed, or the NS decoding subtasks are not completed, the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers, and the NS decoding subtasks are the decoding subtasks among the N decoding subtasks except the S decoding subtasks.

[0027] In one possible design, the K neural network layers are determined based on capability information and / or channel state information of the terminal device, wherein the capability information of the terminal device is used to indicate the neural network computing capability of the terminal device.

[0028] In a fourth aspect, the present application provides a communication device, comprising: a processing unit and a transceiver unit; the transceiver unit is used to receive a second data packet and a first indication information, the data carried by the second data packet can be decoded by MK neural network layers, the K neural network layers belong to M neural network layers, M and K are positive integers, K<M; the first indication information indicates that it has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers; the processing unit is used to decode the data carried by the second data packet according to the first indication information to obtain business data.

[0029] In one possible design, the transceiver unit is used to receive fourth indication information, where the fourth indication information indicates a mapping relationship between M neural network layers and N decoding subtasks, wherein each of the N decoding subtasks corresponds to at least one neural network layer of the M neural network layers, and N is a positive integer.

[0030] In one possible design, the first indication information indicates that S decoding subtasks have been completed, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, S<N, or the first indication information indicates that NS decoding subtasks are not completed, and the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers.

[0031] In a fifth aspect, the present application provides a communication system, which includes a server, a core network network element and a terminal device; the server is used to generate a first data packet and send the first data packet to the core network network element, wherein the data carried by the first data packet can be decoded by M neural network layers, M is a positive integer; the core network network element is used to receive the first data packet from the server, and send the second data packet and a first indication information; the second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, wherein the K neural network layers belong to the M neural network layers, and the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M; the first indication information indicates that it has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers; the terminal device is used to receive the second data packet and the first indication information from the core network network element, and decode the data carried by the second data packet according to the first indication information to obtain service data.

[0032] In a sixth aspect, the present application provides a communication system, which includes a server, a core network network element, an access network device and a terminal device; the server is used to generate a third data packet and send the third data packet to the core network network element, wherein the data carried by the third data packet can be decoded by M neural network layers, and M is a positive integer; the core network network element is used to receive the third data packet from the server, generate a first data packet based on the third data packet, the data carried by the first data packet is the same as the data carried by the third data packet, and the data carried by the first data packet can be decoded by the M neural network layers; the access network device is used to receive the third data packet from the core network network element. A data packet, and sending the second data packet and the first indication information; the second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, wherein the K neural network layers belong to the M neural network layers, and the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M; the first indication information indicates that it has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers; the terminal device is used to receive the second data packet and the first indication information from the access network device, and decode the data carried by the second data packet according to the first indication information to obtain service data.

[0033] In the seventh aspect, the present application provides a communication device, which can be a first device, or a module or unit (for example, a chip, or a chip system, or a circuit) in the first device that corresponds one-to-one to the method / operation / step / action described in any one of the first to second aspects, or can be used in combination with the first device.

[0034] In an eighth aspect, the present application provides a communication device comprising at least one processing element and at least one storage element, wherein the at least one storage element is used to store programs and data, and the at least one processing element is used to read and execute the programs and data stored in the storage element, so that any method described in any one of the above aspects of the present application is implemented.

[0035] In a ninth aspect, the present application further provides a computer program, which, when executed on a computer, enables the computer to execute any of the methods described in any of the above aspects.

[0036] In the tenth aspect, the present application provides a communication device, which includes: an interface circuit and at least one processor; the interface circuit is used to provide input and / or output of programs or instructions to the at least one processor; the at least one processor is used to execute the program or instructions so that the communication device can implement any method described in any of the above aspects.

[0037] In one possible manner, the communication device includes the at least one memory, and the at least one memory is used to store the program or instruction.

[0038] In an eleventh aspect, the present application provides a computer storage medium storing a software program. When the software program is read and executed by one or more processors, the software program can implement any of the methods described in any of the above aspects.

[0039] In a twelfth aspect, the present application provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute any of the methods described in any of the above aspects.

[0040] In a thirteenth aspect, the present application provides a chip system, which includes at least one chip and a memory, and the at least one chip is used to read and execute a program stored in the memory to implement any of the methods described in any of the above aspects.

[0041] Based on the implementations provided in the above aspects, this application can also be further combined to provide more implementations. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] FIG1 shows a schematic diagram of the transmission process of an XR video frame;

[0043] FIG2 shows an architecture diagram of a communication system;

[0044] FIG3A shows a schematic diagram of a system architecture 1 that may be used in the present application;

[0045] FIG3B shows a schematic diagram of a system architecture 2 that may be used in the present application;

[0046] FIG3C shows a schematic diagram of a system architecture 3 that may be used in the present application;

[0047] FIG4 shows a possible flow diagram of a communication method provided in an embodiment of the present application;

[0048] FIG5A shows a schematic diagram of a data packet transmission and decoding process;

[0049] FIG5B shows one of the detailed flow charts of data packet transmission and decoding;

[0050] FIG6A shows a second schematic diagram of the data packet transmission and decoding process;

[0051] FIG6B shows a second detailed flow chart of data packet transmission and decoding;

[0052] FIG7 is a schematic structural diagram of a communication device in this application;

[0053] FIG8 is a schematic structural diagram of another communication device in this application. DETAILED DESCRIPTION

[0054] The specific implementation of the present application is described below with reference to the accompanying drawings in the embodiments of the present application. However, the implementation of the present application may also include combining these embodiments without departing from the spirit or scope of the present application, such as adopting other embodiments and making structural changes. Therefore, the detailed description of the following embodiments should not be understood in a restrictive sense. The terms used in the examples section of the present application are only used to explain the specific embodiments of the present application and are not intended to limit the present application.

[0055] In recent years, with the continuous development of the fifth-generation (5G) communication system, data transmission latency has been continuously reduced and transmission capacity has been increasing. 5G communication systems have gradually penetrated into some multimedia services with strong real-time requirements and large data capacity, such as video transmission, cloud gaming (CG), and XR, among which XR includes virtual reality (VR) and augmented reality (AR).

[0056] With the rapid increase in communication transmission rates, real-time video transmission has gradually become one of the core services in current networks. The continuous advancement and improvement of extended reality technology has also led to the rapid development of related industries. XR technology has now entered various fields closely related to people's production and daily life, such as education, entertainment, military, medical care, environmental protection, transportation, and public health. Compared with traditional video services, XR offers advantages such as multiple perspectives and strong interactivity, providing users with a brand new visual experience.

[0057] The core concept of the AI ​​feature stream-based encoding and decoding method is to solve the expression and transmission of information meaning at the feature stream level (semantic level), partially or completely moving the understanding of the information meaning to the sending end, thereby reducing the transmission volume and the demand for transmission bandwidth. In the current XR service transmission architecture, the media encoding and decoding functions of the source are performed by the cloud server and the terminal device respectively, and the network is only responsible for the transmission function. The communication architecture using the AI ​​feature stream encoding and decoding method is shown in Figure 1.

[0058] As shown in Figure 1, the server encodes the business data of the XR video frame through AI feature stream encoding (or AI feature stream encoding neural network) and assembles it into multiple Internet Protocol (IP) packets, such as 50 IP packets, which are then transmitted through the fixed network / core network to the radio access network (RAN), that is, the base station side, and then transmitted to the terminal device through the wireless air interface. The terminal device decodes the encoded data in the IP packet through the AI ​​feature stream (or AI feature stream decoding neural network) to restore the business data of the XR video frame and display it through the display device. Table 1 shows the types of neural network layers in the AI ​​feature stream encoding and decoding neural network and their related functions and characteristics.

[0059] Table 1

[0060] It should be noted that the encoding and decoding methods involved in the embodiments of the present application refer to encoding and decoding through a neural network, wherein an encoding neural network can be used to encode the service data to obtain the encoded data, and a decoding neural network can be used to decode the encoded data to obtain the service data. The encoding and decoding methods involved in the embodiments of the present application include but are not limited to the encoding and decoding methods of the AI ​​feature stream. The following content is only explained by taking the AI ​​feature stream encoding method as an example.

[0061] Exemplarily, the encoding neural network can be referred to as the encoding network or editor. Exemplarily, the encoding neural network can be an AI feature stream encoding neural network, or an AI encoder, etc., which is not limited in this application. Generally, the encoding neural network includes multiple neural network layers. Exemplarily, the front end of the encoding neural network is usually a convolutional layer and a pooling layer, the purpose of which is to extract the feature information of the data frame and reduce the redundancy between the business data carried by the data frame. The back end is usually a fully connected layer and an activation layer, wherein the fully connected layer is used to further classify the extracted feature information. The role of the activation layer is to introduce nonlinearity, and through nonlinearity, the neural network approximates any function, thereby solving complex real-world problems.

[0062] The decoding neural network can be referred to as a decoding network or decoder. For example, the decoding neural network can be an AI feature stream decoding neural network, or an AI decoder, etc., which is not limited in this application. Generally, the decoding neural network includes multiple neural network layers. The role of the decoding neural network is opposite to that of the encoding neural network. The structure of the decoding neural network is symmetrical to that of the encoding neural network. The front end is the fully connected layer and the activation layer, and the back end is the convolution layer and the pooling layer. Generally speaking, the closer to the end of the decoder, the greater the amount of data generated.

[0063] It should be noted that the number of neural network layers included in the encoding neural network and the number of neural network layers included in the decoding neural network can be the same or different, and this application does not limit this.

[0064] The embodiments of the present application can be applied to various communication systems, such as: long term evolution (LTE) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD), fifth generation (5G) system or new radio (NR), or applied to future communication systems or other similar communication systems.

[0065] Figure 2 shows the architecture of the 5G communication system, as standardized by the 3rd Generation Partnership Project (3GPP). The 5G network architecture shown in Figure 2 includes terminal devices, access network (AN) equipment, and core network elements. Terminal devices access the data network (DN) through the access network equipment and core network elements.

[0066] The access network device may be a radio access network (RAN) device. For example: a base station, an evolved NodeB (eNodeB), a transmission reception point (TRP), a next generation NodeB (gNB) in a 5G mobile communication system, a next generation base station in a sixth generation (6G) mobile communication system, a base station in a future mobile communication system, or an access node in a wireless fidelity (WiFi) system, etc.; it may also be a module or unit that completes part of the functions of a base station, for example, a centralized unit (CU) or a distributed unit (DU). The radio access network device may be a macro base station, a micro base station or an indoor station, a relay node or a donor node, etc. The embodiments of the present application do not limit the specific technology and specific device form adopted by the access network device.

[0067] Terminal devices can be user equipment (UE), mobile stations, mobile terminals, etc. Terminal devices can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), the Internet of Things (IoT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grids, smart furniture, smart offices, smart wearables, smart transportation, and smart cities. Terminal devices can be mobile phones, tablets, computers with wireless transceiver capabilities, wearable devices, vehicles, urban air vehicles (such as drones and helicopters), ships, robots, robotic arms, smart home devices, etc.

[0068] Core network elements include user plane function (UPF) elements, access and mobility management function (AMF) elements, session management function (SMF) elements, network exposure function (NEF) elements, network function repository function (NRF) elements, unified data management (UDM) elements, policy control function (PCF) elements, and application function (AF) elements. The UPF element is a user plane element, and the other elements are control plane elements.

[0069] The interface between each control plane network element can be a service-oriented interface (as shown in Figure 2) or a point-to-point interface. This application does not limit this and only uses Figure 2 as an example for explanation.

[0070] The following is a brief introduction to some core network equipment:

[0071] 1. The SMF network element (SMF) is primarily used for session management, terminal device IP address allocation and management, selection of endpoints for manageable user equipment plane functions, policy control, or charging function interfaces, and downlink data notification. Nsmf is a service-based interface provided by SMF, through which SMF can communicate with other network functions.

[0072] 2. AMF network element, referred to as AMF, is mainly used for mobility management and access management. NMF is a service-based interface provided by AMF. AMF can communicate with other network functions through NMF.

[0073] 3. The UDM network element, or UDM for short, handles user identification, subscription, access authentication, registration, or mobility management. Nudm is the service-based interface provided by the UDM, through which the UDM can communicate with other network functions.

[0074] 4. UPF network element, referred to as UPF, is used for packet routing and forwarding, or quality of service (QoS) processing of user plane data.

[0075] 5. NEF network element, referred to as NEF, is used to expose the services and capabilities of 3GPP network functions to AF, and also allows AF to provide information to 3GPP network functions.

[0076] 6. PCF network element, referred to as PCF, is used for policy management of charging policy and QoS policy.

[0077] It is understandable that the core network network elements may also include other network elements, and this application does not limit this. The above network elements are examples of one implementation method, and this application does not exclude the existence of network elements or devices with the above network element functions in 6G or newer wireless communication systems with other names or other forms. The above network elements or functions can be network elements in hardware devices, software functions running on dedicated hardware, or virtualized functions instantiated on a platform (for example, a cloud platform). As a possible implementation method, the above network elements or functions can be implemented by one device, or by multiple devices together, or as a functional module within a device, and this embodiment of the application does not specifically limit this.

[0078] The following describes the system architecture that may be applied to this application. It should be understood that the following architectures are only examples, and the names of specific network element nodes are also only examples.

[0079] System Architecture 1: Server-Network-UE Architecture

[0080] As shown in Figure 3A, in system architecture 1, a server can transmit data to a user end user (UE) via a network. For example, the server can implement video source encoding, decoding, and rendering. The network can include, but is not limited to, the following devices: a network device (e.g., a fixed network), core network elements (e.g., a UPF), and access network equipment. The user end user (UE) can include devices such as head-mounted XR glasses, video players, and holographic projectors.

[0081] System Architecture 2: UE-Network-UE Architecture

[0082] As shown in Figure 3B, in system architecture 2, UE1 can perform data transmission with UE2 via the network. Specific devices that may be included in the network can refer to architecture 1.

[0083] System Architecture 3: WiFi Scenario

[0084] As shown in FIG3C , in system architecture 3 , the server can transmit data with the UE via a fixed network, a WiFi router, an access point (AP), or a set-top box.

[0085] Based on the above system architecture and the contents of the above related technical introduction, a possible communication method is provided in an embodiment of the present application, and the execution subjects of each communication method are introduced by taking the first device and the terminal device as examples. For example, the first device can be the access network device or the core network element in Figures 3A and 3B above, or the first device can be the fixed network, WiFi router, AP or set-top box in Figure 3C above. The terminal device can be any UE shown in Figures 3A to 3C above. In addition, it should be understood that the first device can also be replaced by a communication device having the function of the first device or a chip, unit or module inside a communication device having the function of the first device. The terminal device can also be replaced by a communication device having the function of a terminal device or a chip, unit or module inside a communication device having the function of a terminal device.

[0086] FIG4 shows a possible flow diagram of a communication method provided in an embodiment of the present application, the method comprising:

[0087] Step 400: A first device receives a first data packet.

[0088] Exemplarily, the data carried by the first data packet can be decoded by M neural network layers, which can also be understood as the data carried by the first data packet can be decoded by M neural network layers. Wherein, M is a positive integer. Exemplarily, the data carried by the first data packet is the data after passing through the encoding neural network, and the data carried by the first data packet can be completely decoded only after passing through all the neural network layers in the decoding neural network to obtain the data before encoding, that is, the business data. Wherein, the decoding neural network includes M neural network layers. The number of neural network layers included in the encoding neural network may be equal to M or may not be equal to M, and this application does not limit this.

[0089] It is understandable that the number of neural network layers included in the decoding neural networks of different services may be different. For example, the decoding neural network of service 1 has 8 layers, and the decoding neural network of service 2 has 9 layers. Exemplarily, if the first data packet is a data packet of the first service, then before the first device receives the first data packet, the first device may receive configuration information indicating that the decoding neural network of the first service includes M neural network layers.

[0090] In one possible implementation, if the first device is a core network element, the first data packet may come from a server, or, if the first device is an access network device, the first data packet may come from a core network element. For details, please refer to the following Figures 5A and 5B, and the embodiments shown in Figures 6A and 6B.

[0091] The second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, that is, the first device decodes the data carried by the first data packet through K neural network layers to obtain the second data packet.

[0092] The data carried by the second data packet can be decoded by MK neural network layers, which can also be understood as the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M. K neural network layers belong to M neural network layers. In other words, the first device can perform partial neural network layer decoding on the data carried by the first data packet to obtain the second data packet. Therefore, the complete decoding of the data carried by the second data packet can also be decoded by MK neural network layers. In other words, the complete decoding of the data carried by the second data packet still needs to be decoded by MK neural network layers.

[0093] Exemplarily, the K neural network layers are the first K neural network layers of the M neural network layers. For example, if M=9 and K=3, the first device performs decoding of the first 3 neural network layers of the 9 neural network layers on the data carried by the first data packet.

[0094] In addition, as a possible implementation method, the first device may also perform decoding of all neural network layers on the data carried by the first data packet, that is, K=M, or it can be understood as completely decoding the data carried by the first data packet, which is not limited in this application.

[0095] Exemplarily, before the first device performs decoding, the first device may also determine whether the data carried by the first data packet can or is capable of being decoded by the neural network in the following manner.

[0096] Method 1: The first data packet may include second indication information, and the second indication information indicates that the data carried by the first data packet can or can be decoded by a neural network. It can also be understood that the second indication information indicates that the data carried by the first data packet can be decoded by a neural network. The neural network here may refer to a decoding neural network. Alternatively, the second indication information may indicate a decoding method for the data carried by the first data packet, such as AI feature stream decoding. Using the above method 1, the first data packet can carry the second indication information, so that the first device knows that the first data packet can or can be decoded by a neural network.

[0097] Method 2: The first device may also receive third indication information, which indicates that the data carried by the first packet is or can be decoded by the neural network. Alternatively, this third indication information may indicate that the data carried by the first packet is or can be decoded by the neural network. Using this method, the sender of the first packet can directly notify the first device through signaling that the data carried by the packet is or can be decoded by the neural network.

[0098] In addition, in a possible implementation, before the first device performs decoding, the first device may obtain the capability information and / or channel state information of the terminal device. For example, the terminal device sends the capability information and / or channel state information of the terminal device to the first device. Then, the first device determines K neural network layers based on the capability information and / or channel state information of the terminal device, and determines the value of K. That is, the K neural network layers are determined based on the capability information and / or channel state information of the terminal device. It can also be understood that the first device can determine whether it is necessary to perform decoding of some neural network layers based on the capability information and / or channel state information of the terminal device, and when it is determined that decoding of some neural network layers needs to be performed, which specific neural network layers need to be decoded.

[0099] Among them, the capability information of the terminal device is used to indicate the neural network computing capability of the terminal device, and the neural network computing capability of the terminal device can also be referred to as the computing power of the terminal device, or the computing power level, or the floating-point computing capability, etc. For example, the capability information of the terminal device includes at least one of the floating-point operations per second (FLOPS), the main frequency of the chip, the memory size, or the time to run certain predefined computing tasks. For example, the channel state information of the terminal device may include the channel state information (CSI) reported by the terminal device, or at least one of the average service rates detected by the quality of service (QoS) flow. It is understandable that the specific content included in the capability information of the above-mentioned terminal device or the specific content included in the channel state information of the terminal device is only an example and is not intended to be limiting of this application. In addition, the capability information of the terminal device or the channel state information of the terminal device may also include other content.

[0100] For example, if the neural network computing capability of the terminal device is relatively poor, the first device can take on more decoding work, that is, the value of K can be relatively large. If the neural network computing capability of the terminal is acceptable, but the channel state information indicates that the channel transmission conditions between the first device and the terminal device are relatively poor, the first device can take on less decoding work, that is, the value of K can be relatively small, and the terminal device can be more responsible for decoding. Since the closer to the end of the decoding neural network, the greater the amount of data generated, the first device can take on less decoding work, which can also reduce the amount of data transmitted on the air interface and reduce transmission delay. If the neural network computing capability of the terminal device is relatively good, the first device may not bear the decoding work, for example, K may also be equal to 0.

[0101] In addition, the first device may also obtain parameters such as the power level or temperature of the terminal device. For example, when the terminal device is unable to maintain a high neural network computing capability due to insufficient power, the first device may assume more decoding work, that is, the value of K may be larger. It is understood that the first device may determine the value of K based on a variety of factors, or the value of K may be configured in advance, and this application does not limit this.

[0102] Step 410: The first device sends a second data packet and first indication information.

[0103] Exemplarily, the first device sends the second data packet and the first indication information to the terminal device, and correspondingly, the terminal device receives the second data packet and the first indication information.

[0104] The first indication information indicates that the decoding has been completed through K neural network layers, or that the decoding has not been completed through MK neural network layers. In other words, the first indication information can indicate a neural network layer that has completed decoding, or a neural network layer that has not completed decoding.

[0105] For example, if M=9 and K=3, the first indication information may indicate that it has been decoded through the first three of the nine neural network layers, or that it has not been decoded through the last six of the nine neural network layers.

[0106] It is understandable that the capability information and / or channel status information of the terminal device may also be updated. The first device can update the value of K based on the updated capability information and / or channel status information of the terminal device, and also update the first indication information.

[0107] For example, assuming M=9, if the neural network computing capability of the terminal is acceptable, but the channel state information indicates that the channel transmission condition between the first device and the terminal device is poor, the first device can determine that the value of K is 2, and the first device performs decoding of the first two of the nine neural network layers on the received data packet. At this time, the first indication information indicates that the decoding has passed the first two of the nine neural network layers. If after a period of time, the second device obtains updated channel state information, and the updated channel state information indicates that the channel transmission condition between the first device and the terminal device has improved, then the first device can adjust the value of K. For example, if the value of K is 4, then the first device performs decoding of the first four of the nine neural network layers on the received data packet, and simultaneously updates the first indication information. The updated first indication information indicates that the decoding has passed the first four of the nine neural network layers.

[0108] In addition, in a possible implementation, the first device may also receive fourth indication information, where the fourth indication information indicates a mapping relationship between M neural network layers and N decoding subtasks, wherein each of the N decoding subtasks corresponds to at least one of the M neural network layers, and N is a positive integer. Exemplarily, a decoding subtask may also be referred to as a subtask, an atomic task, or a decoding atomic task, etc. The specific meaning of the mapping relationship between M neural network layers and N decoding subtasks can be understood as defining a mapping relationship between subtasks that require actual calculation or execution and decoding neural network layers. This application does not limit this. The N decoding subtasks are different, and the neural network layers included in different decoding subtasks are not repeated. The number of neural network layers included in different decoding subtasks may be the same or different. The mapping relationship between M neural network layers and N decoding subtasks may be determined by the server and notified to the first device.

[0109] For example, if M=9 and N=3, the mapping relationship between the 9 neural network layers and the 3 decoding subtasks can be shown in Table 2 below.

[0110] Table 2

[0111] Furthermore, the first device may decode the data carried by the first data packet through K neural network layers according to the fourth indication information. Exemplarily, the first device may perform S decoding subtasks out of N decoding subtasks on the data carried by the first data packet, where the neural network layers corresponding to the S decoding subtasks are K neural network layers, where S is a positive integer and S<N.

[0112] Exemplarily, the S decoding subtasks among the N decoding subtasks may be understood as the first S decoding subtasks among the N decoding subtasks.

[0113] Correspondingly, the first indication information indicates that S decoding subtasks have been completed, or NS decoding subtasks have not been completed, the neural network layers corresponding to the NS decoding subtasks are MK neural network layers, and the NS decoding subtasks are the decoding subtasks among the N decoding subtasks except the S decoding subtasks.

[0114] For example, the first device can perform the first decoding subtask of the three decoding subtasks on the data carried by the first data packet according to the above Table 2, that is, decode through the first three neural network layers to obtain the second data packet, and send the second data packet and the first indication information to the terminal device, wherein the first indication information indicates that the first decoding subtask has been completed, or that the second decoding subtask and the third decoding subtask are not completed.

[0115] Step 420: The terminal device decodes the data carried by the second data packet according to the first indication information to obtain service data.

[0116] In one possible implementation, the terminal device determines which neural network layers still need to be decoded for the second data packet based on the first indication information, and completes the decoding of these neural network layers to obtain service data, or the service delay may also be considered. When the service delay is long, the terminal device may no longer continue to decode the data. For example, the first indication information indicates that it has been decoded through K neural network layers, or has not been decoded through MK neural network layers. Then the terminal device determines that the data carried by the second data packet can also be decoded through MK neural network layers based on the first indication information. That is, the terminal device determines that the data carried by the second data packet still needs to be decoded through MK neural network layers based on the first indication information, and then the terminal device decodes the data carried by the second data packet through MK neural network layers to obtain service data. Therefore, the first device and the terminal device can jointly complete the decoding of the data carried by the first data packet, wherein the first device performs decoding of K neural network layers, and the terminal device performs decoding of MK neural network layers.

[0117] In addition, in a possible implementation, the terminal device may also receive fourth indication information, and then the terminal device may decode the data carried by the second data packet according to the fourth indication information and the first indication information.

[0118] For example, in combination with Table 2 above, the first device sends a second data packet and a first indication message to the terminal device, wherein the first indication message indicates that the first decoding subtask has been completed, or that the second decoding subtask and the third decoding subtask have not been completed. At this time, the terminal device can determine that the data carried by the second data packet still needs to execute the second decoding subtask and the third decoding subtask based on the fourth indication message (such as Table 2) and the first indication message, that is, the data carried by the second data packet can or can be decoded by the 4th to 9th neural network layers, and then the data carried by the second data packet is decoded by the 4th to 9th neural network layers.

[0119] Using the above method, the first device can first perform decoding of part of the neural network layer, and the terminal device can perform decoding of the remaining neural network layers, thereby reducing the dependence on the neural network computing capability of the terminal device and saving the power consumption and cost of the terminal device. Compared with the terminal device, the first device has stronger neural network computing capability, which can improve decoding efficiency, reduce the transmission delay of data packets, and improve user service experience.

[0120] The method embodiment shown in FIG4 is further described below with reference to specific embodiments:

[0121] FIG. 5A and FIG. 5B show one of the specific flow charts of data packet transmission and decoding.

[0122] S501: The server generates a first data packet.

[0123] Exemplarily, the server encodes the data of the first service through an encoding neural network to generate a first data packet, wherein the data carried by the first data packet can or can be decoded by M neural network layers, M is a positive integer, and the first data packet is a data packet of the first service.

[0124] In addition, before the first service starts, for example, before S501, the server can determine that the decoding neural network of the first service includes M neural network layers, and determine the decoding subtask set based on the neural network layers included in the decoding neural network of the first service, that is, determine the mapping relationship between the M neural network layer decodings and the N decoding subtasks. Further, the server can notify the core network device and the terminal device of the mapping relationship between the M neural network layer decodings and the N decoding subtasks. For example, the server can notify the core network network element of the mapping relationship between the M neural network layer decodings and the N decoding subtasks through the N33 interface, and the core network network element can notify the terminal device of the mapping relationship between the M neural network layer decodings and the N decoding subtasks through non-access stratum (NAS) signaling.

[0125] For example, the decoding neural network of the first service includes 9 neural network layers. The server determines 3 decoding subtasks based on the 9 neural network layers. The specific mapping relationship can be shown in Table 2.

[0126] S502: The server sends a first data packet to a core network element. Correspondingly, the core network element receives the first data packet from the server.

[0127] Exemplarily, the server may carry the second indication information in the first data packet, that is, add the second indication information to the first data packet. Furthermore, the core network element may determine that the received first data packet can or can be decoded using a neural network based on the second indication information.

[0128] For example, the server may add the second indication information to a real-time transport protocol (RTP) header of the first data packet.

[0129] S503: The core network element decodes the data carried by the first data packet through K neural network layers to obtain a second data packet.

[0130] Exemplarily, after determining that the received first data packet can or can be decoded using a neural network, and before decoding the data carried by the first data packet, the core network element can determine which neural network layers need to be decoded based on the obtained capability information and / or channel status of the terminal device, for example, determine which decoding subtasks of N decoding subtasks to execute. For details, please refer to the relevant description in the above step 410, which will not be repeated here.

[0131] S504: The core network element sends the second data packet and the first indication information to the terminal device. Correspondingly, the terminal device receives the second data packet and the first indication information from the core network element.

[0132] The data carried by the second data packet can or can be decoded by MK neural network layers, K is a positive integer, K<M, and the first indication information indicates that it has been decoded by K neural network layers, or has not been decoded by MK neural network layers.

[0133] Exemplarily, the core network element may carry the first indication information via NAS signaling.

[0134] S505: The terminal device decodes the data carried by the second data packet according to the first indication information to obtain service data.

[0135] Exemplarily, the terminal device determines the neural network layer that has not completed decoding based on the first indication information, decodes the data carried by the second data packet, and obtains service data.

[0136] Using the above method, the core network network element can first perform decoding of part of the neural network layer, and the terminal device performs decoding of the remaining neural network layers, thereby reducing the dependence on the neural network computing capabilities of the terminal device and saving the power consumption and cost of the terminal device. Compared with the terminal device, the core network network element has stronger neural network computing capabilities, which can improve the decoding efficiency, reduce the transmission delay of the data packet, and improve the user service experience.

[0137] FIG. 6A and FIG. 6B show one of the specific flow charts of data packet transmission and decoding.

[0138] S601: The server generates a third data packet.

[0139] Exemplarily, the server encodes the data of the first service through an encoding neural network to generate a third data packet, wherein the data carried by the third data packet can or can be decoded by M neural network layers, M is a positive integer, and the third data packet is the data packet of the first service.

[0140] In addition, before the first service starts, for example, before S601, the server can determine that the decoding neural network of the first service includes M neural network layers, and determine a set of decoding subtasks based on the neural network layers included in the decoding neural network of the first service, that is, determine the mapping relationship between the M neural network layer decoding and the N decoding subtasks. Further, the server can notify the access network device and the terminal device of the mapping relationship between the M neural network layer decoding and the N decoding subtasks. For example, the server can notify the core network element of the mapping relationship between the M neural network layer decoding and the N decoding subtasks through the N33 interface, and the core network element can notify the access network device of the mapping relationship between the M neural network layer decoding and the N decoding subtasks through general packet radio system tunneling protocol control plane (GTP-C) signaling. The core network element can notify the terminal device of the mapping relationship between the M neural network layer decoding and the N decoding subtasks through NAS signaling, or the access network device can notify the terminal device of the mapping relationship between the M neural network layer decoding and the N decoding subtasks through a media access control control element (MAC CE).

[0141] For example, the decoding neural network of the first service includes 9 neural network layers. The server determines 3 decoding subtasks based on the 9 neural network layers. The specific mapping relationship is shown in Table 2.

[0142] S602: The server sends a third data packet to the core network element. Correspondingly, the core network element receives the third data packet from the server.

[0143] Exemplarily, the server may carry the second indication information via a third data packet, that is, add the second indication information to the first data packet. Furthermore, the core network element may determine, based on the second indication information, that the received third data packet can or can be decoded using a neural network.

[0144] For example, the server may add the second indication information to the RTP header of the third data packet.

[0145] S603: The core network element generates a first data packet based on the third data packet. The data carried by the first data packet is the same as the data carried by the third data packet. The data carried by the first data packet can be or can be decoded by M neural network layers.

[0146] For example, the core network element may encapsulate the third data packet using a general packet radio system tunneling protocol-user plane (GTP-U) to obtain the first data packet. The GTP header of the first data packet carries second indication information, and the second indication information indicates that the data carried by the first data packet can be or is capable of being decoded by the neural network.

[0147] Alternatively, another core network element (e.g., SMF) may send a notification message to the access network device, the notification message including the second indication information. In this case, the core network element (e.g., UPF) may not require the GTP header to carry the second indication information. The second indication information indicates that the data carried by the first packet can or can be decoded by the neural network. For example, the core network element (e.g., SMF) may add the second indication information to the QoS description information (profile).

[0148] S604: The core network element sends a first data packet to the access network device. Correspondingly, the access network device receives the first data packet from the core network element.

[0149] S605: The access network device decodes the data carried by the first data packet through K neural network layers to obtain a second data packet.

[0150] Exemplarily, after determining that the received first data packet can or is capable of being decoded using a neural network, and before decoding the data carried by the first data packet, the access network device can determine which neural network layers need to be decoded based on the obtained capability information and / or channel status of the terminal device, for example, determine which decoding subtasks of N decoding subtasks to execute. For details, please refer to the relevant description in the above step 410, which will not be repeated here.

[0151] S606: The access network device sends the second data packet and the first indication information. Correspondingly, the terminal device receives the second data packet and the first indication information from the access network device.

[0152] The data carried by the second data packet can or can be decoded by MK neural network layers, where K is a positive integer and K<M; the first indication information indicates that the data has been decoded by K neural network layers or has not been decoded by MK neural network layers.

[0153] For example, the first indication information may be carried by a MAC CE.

[0154] S607: The terminal device decodes the data carried by the second data packet according to the first indication information to obtain service data.

[0155] Exemplarily, the terminal device determines the neural network layer that has not completed decoding based on the first indication information, decodes the data carried by the second data packet, and obtains service data.

[0156] Using the above method, the access network device can first perform decoding of part of the neural network layer, and the terminal device can perform decoding of the remaining neural network layers, thereby reducing the dependence on the neural network computing capabilities of the terminal device and saving the power consumption and cost of the terminal device. Compared with the terminal device, the access network device has stronger neural network computing capabilities, which can improve the decoding efficiency, reduce the transmission delay of the data packet, and improve the user service experience.

[0157] It is understandable that in order to implement the functions in the above embodiments, the terminal device and the first device include hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily appreciate that, in combination with the units and method steps of each example described in the embodiments disclosed in this application, this application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in hardware or in a manner driven by computer software depends on the specific application scenario and design constraints of the technical solution.

[0158] Figures 7 and 8 are schematic diagrams of possible communication devices provided in embodiments of the present application. These communication devices can be used to implement the functions of the terminal device or the first device in the above method embodiments, and thus can also achieve the beneficial effects of the above method embodiments.

[0159] As shown in Figure 7, a communication device 700 includes a processing unit 710 and a transceiver unit 720. The communication device 700 is used to implement the terminal device or the first device in the above method embodiment.

[0160] When the communication device 700 is used to implement the function of the first device in the method embodiment shown in FIG4 :

[0161] The transceiver unit 720 is used to send and receive information;

[0162] The processing unit 710 is used to receive a first data packet through the transceiver unit 720; wherein the data carried by the first data packet can be decoded by M neural network layers, M is a positive integer; and send the second data packet and the first indication information through the transceiver unit 720, wherein the second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, and the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M; the K neural network layers belong to the M neural network layers; the first indication information indicates that it has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers.

[0163] In one possible design, the first data packet includes second indication information, where the second indication information indicates that data carried by the first data packet can be decoded by a neural network.

[0164] In one possible design, the transceiver unit 720 is used to receive third indication information, where the third indication information indicates that the data carried by the first data packet can be decoded by a neural network.

[0165] In one possible design, the transceiver unit 720 is used to receive fourth indication information, where the fourth indication information indicates a mapping relationship between the M neural network layers and the N decoding subtasks, wherein each of the N decoding subtasks corresponds to at least one of the M neural network layers, and N is a positive integer; the second data packet is obtained by executing S of the N decoding subtasks on the data carried by the first data packet, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, and S<N.

[0166] In one possible design, the first indication information indicates that the S decoding subtasks have been completed, or the NS decoding subtasks are not completed, the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers, and the NS decoding subtasks are the decoding subtasks among the N decoding subtasks except the S decoding subtasks.

[0167] In one possible design, the K neural network layers are determined based on capability information and / or channel state information of the terminal device, wherein the capability information of the terminal device is used to indicate the neural network computing capability of the terminal device.

[0168] When the communication device 700 is used to implement the functions of the terminal device in the method embodiment shown in FIG4 :

[0169] The transceiver unit 720 is configured to receive a second data packet and first indication information, where the data carried by the second data packet can be decoded by MK neural network layers, where the K neural network layers belong to M neural network layers, M and K are positive integers, and K<M; and the first indication information indicates that the data has been decoded by the K neural network layers or has not been decoded by the MK neural network layers.

[0170] The processing unit 710 is configured to decode the data carried by the second data packet according to the first indication information to obtain service data.

[0171] In one possible design, the transceiver unit 720 is used to receive fourth indication information, where the fourth indication information indicates a mapping relationship between M neural network layers and N decoding subtasks, wherein each of the N decoding subtasks corresponds to at least one neural network layer of the M neural network layers, and N is a positive integer.

[0172] In one possible design, the first indication information indicates that S decoding subtasks have been completed, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, S<N, or the first indication information indicates that NS decoding subtasks are not completed, and the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers.

[0173] A more detailed description of the processing unit 710 and the transceiver unit 720 can be directly obtained by referring to the relevant description in the above method embodiment, and will not be repeated here.

[0174] As shown in Figure 8, communication device 800 includes a processor 810 and an interface circuit 820. Processor 810 and interface circuit 820 are coupled to each other. It will be appreciated that interface circuit 820 may be a transceiver or an input / output interface. Optionally, communication device 800 may further include a memory 830 for storing instructions executed by processor 810, input data required by processor 810 to execute instructions, or data generated after processor 810 executes instructions.

[0175] When the communication device 800 is used to implement the method shown in FIG. 4 , the processor 810 is used to implement the functions of the processing unit 710 , and the interface circuit 820 is used to implement the functions of the transceiver unit 720 .

[0176] It is understood that the processor 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, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0177] This application provides another example of a device, which includes at least one processor and at least one memory, the at least one processor and the at least one memory being coupled together, the at least one memory being used to store instructions. When the instructions are executed by the at least one processor, the communication device performs the method in the above-described embodiment. For example, as shown in FIG8 , a communication device 800 includes a processor 810 and a memory 830. The processor 810 and the memory 830 are coupled together, and the memory 830 stores instructions. When the instructions stored in the memory 830 are executed by the processor 810, the communication device 800 performs the method performed by the terminal device or the first device in the above-described embodiment.

[0178] The method steps in the embodiments of the present application can be implemented in hardware or in software instructions that can be executed by a processor. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disk, mobile hard disk, CD-ROM or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. The storage medium can also be an integral part of the processor. The processor and storage medium can be located in an ASIC. In addition, the ASIC can be located in the above-mentioned terminal device or the first device. The processor and storage medium can also exist in the terminal device or the first device as discrete components.

[0179] In the above embodiments, all or part of the embodiments may be implemented using 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 programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are performed in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, or other programmable device. The computer program or instructions may be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions may be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a digital video disk; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or nonvolatile storage medium, or may include both volatile and nonvolatile types of storage media.

[0180] In the various embodiments of the present application, unless otherwise specified or there is any logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced by each other. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0181] In this application, "at least one" means one or more, and "more" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. In the text description of this application, the character " / " generally indicates that the previous and next associated objects are in an "or" relationship; in the formula of this application, the character " / " indicates that the previous and next associated objects are in a "division" relationship. "Including at least one of A, B and C" can mean: including A; including B; including C; including A and B; including A and C; including B and C; including A, B and C.

[0182] It is understood that the various numbers used in the embodiments of this application are merely for ease of description and are not intended to limit the scope of the embodiments of this application. The order of the sequence numbers of the above-mentioned processes does not necessarily imply a specific order of execution; the order of execution of the processes should be determined by their functions and inherent logic.

Claims

1. A communication method, characterized in that: The method includes: Receive a first data packet; wherein the data carried by the first data packet can be decoded by M neural network layers, where M is a positive integer; A second data packet and a first indication message are sent, wherein the second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, and the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M; the K neural network layers belong to the M neural network layers, and the first indication information indicates that the data has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers.

2. The method according to claim 1, characterized in that The first data packet includes second indication information, and the second indication information indicates that the data carried by the first data packet can be decoded by a neural network.

3. The method according to claim 1, characterized in that Also includes: Receive third indication information, where the third indication information indicates that data carried by the first data packet can be decoded by a neural network.

4. The method according to any one of claims 1 to 3, characterized in that: Also includes: Receive fourth indication information, where the fourth indication information indicates a mapping relationship between the M neural network layers and the N decoding subtasks, wherein each decoding subtask in the N decoding subtasks corresponds to at least one neural network layer in the M neural network layers, and N is a positive integer; The second data packet is obtained by executing S decoding subtasks among the N decoding subtasks on the data carried by the first data packet, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, and S<N.

5. The method according to claim 4, characterized in that The first indication information indicates that the S decoding subtasks have been completed, or that the NS decoding subtasks have not been completed, the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers, and the NS decoding subtasks are the decoding subtasks among the N decoding subtasks excluding the S decoding subtasks.

6. The method according to any one of claims 1 to 5, characterized in that: The K neural network layers are determined based on capability information and / or channel state information of the terminal device, wherein the capability information of the terminal device is used to indicate the neural network computing capability of the terminal device.

7. A communication method, characterized in that: The method includes: receiving a second data packet and first indication information, wherein the data carried by the second data packet can be decoded by MK neural network layers, the K neural network layers belong to M neural network layers, the M and K are positive integers, K<M; and the first indication information indicates that the data has been decoded by the K neural network layers or has not been decoded by the MK neural network layers; The data carried by the second data packet is decoded according to the first indication information to obtain service data.

8. The method according to claim 7, characterized in that Also includes: Receive fourth indication information, wherein the fourth indication information indicates a mapping relationship between the M neural network layers and the N decoding subtasks, wherein each decoding subtask of the N decoding subtasks corresponds to at least one neural network layer of the M neural network layers, and N is a positive integer.

9. The method according to claim 8, characterized in that The first indication information indicates that S decoding subtasks have been completed, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, S<N, or the first indication information indicates that NS decoding subtasks are not completed, and the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers.

10. A communication device, characterized in that: The device comprises a processing unit and a transceiver unit; The transceiver unit is used to send and receive information; The processing unit is used to receive a first data packet through the transceiver unit; wherein the data carried by the first data packet can be decoded by M neural network layers, M is a positive integer; and send the second data packet and first indication information through the transceiver unit, wherein the second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, and the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M; the K neural network layers belong to the M neural network layers, and the first indication information indicates that it has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers.

11. The device according to claim 10, characterized in that The first data packet includes second indication information, and the second indication information indicates that the data carried by the first data packet can be decoded by a neural network.

12. The device according to claim 10, characterized in that The transceiver unit is used to receive third indication information, and the third indication information indicates that the data carried by the first data packet can be decoded by the neural network.

13. The device according to any one of claims 10 to 12, characterized in that: The transceiver unit is used to receive fourth indication information, wherein the fourth indication information indicates a mapping relationship between the M neural network layers and the N decoding subtasks, wherein each decoding subtask in the N decoding subtasks corresponds to at least one neural network layer in the M neural network layers, and N is a positive integer; The second data packet is obtained by executing S decoding subtasks among the N decoding subtasks on the data carried by the first data packet, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, and S<N.

14. The device according to claim 13, characterized in that The first indication information indicates that the S decoding subtasks have been completed, or that the NS decoding subtasks have not been completed, the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers, and the NS decoding subtasks are the decoding subtasks among the N decoding subtasks excluding the S decoding subtasks.

15. The device according to any one of claims 10 to 14, characterized in that: The K neural network layers are determined based on capability information and / or channel state information of the terminal device, wherein the capability information of the terminal device is used to indicate the neural network computing capability of the terminal device.

16. A communication device, characterized in that: The device includes: A transceiver unit, configured to receive a second data packet and first indication information, wherein the data carried by the second data packet can be decoded by MK neural network layers, the K neural network layers belong to M neural network layers, the M and K are positive integers, K<M; and the first indication information indicates that the data has been decoded by the K neural network layers or has not been decoded by the MK neural network layers; A processing unit is used to decode the data carried by the second data packet according to the first indication information to obtain business data.

17. The device according to claim 16, characterized in that The transceiver unit is used to receive fourth indication information, where the fourth indication information indicates a mapping relationship between M neural network layers and N decoding subtasks, wherein each decoding subtask of the N decoding subtasks corresponds to at least one neural network layer of the M neural network layers, and N is a positive integer.

18. The device according to claim 17, characterized in that The first indication information indicates that S decoding subtasks have been completed, and the neural network layers corresponding to the S decoding subtasks are the K neural network layers, S is a positive integer, S<N, or the first indication information indicates that NS decoding subtasks are not completed, and the neural network layers corresponding to the NS decoding subtasks are the MK neural network layers.

19. A communication system, characterized in that: The system includes a server, a core network element and a terminal device; The server is configured to generate a first data packet and send the first data packet to the core network element, wherein the data carried by the first data packet can be decoded by M neural network layers, where M is a positive integer; The core network element is used to receive the first data packet from the server, and send the second data packet and first indication information; the second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, wherein the K neural network layers belong to the M neural network layers, and the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M; the first indication information indicates that it has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers; The terminal device is used to receive the second data packet and the first indication information from the core network element, and decode the data carried by the second data packet according to the first indication information to obtain service data.

20. A communication system, characterized in that: The system includes a server, a core network element, an access network device and a terminal device; The server is configured to generate a third data packet and send the third data packet to the core network element, wherein the data carried by the third data packet can be decoded by M neural network layers, where M is a positive integer; The core network element is configured to receive the third data packet from the server, generate a first data packet based on the third data packet, the data carried by the first data packet is the same as the data carried by the third data packet, and the data carried by the first data packet can be decoded by the M neural network layers; The access network device is used to receive the first data packet from the core network network element, and send the second data packet and first indication information; the second data packet is obtained by decoding the data carried by the first data packet through K neural network layers, wherein the K neural network layers belong to the M neural network layers, and the data carried by the second data packet can be decoded by MK neural network layers, K is a positive integer, K<M; the first indication information indicates that it has been decoded by the K neural network layers, or has not been decoded by the MK neural network layers; The terminal device is used to receive the second data packet and the first indication information from the access network device, and decode the data carried by the second data packet according to the first indication information to obtain service data.

21. A communication device, characterized in that: The communication device comprises at least one processor; the at least one processor is configured to execute the method according to any one of claims 1 to 9.

22. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a program, and when the program is run on a device, the device is caused to perform the method according to any one of claims 1 to 9.

23. A computer program product, characterized in that The computer program product comprises a program or instructions, and when the program or instructions are executed by a device, the device is caused to perform the method according to any one of claims 1 to 9.

Images