Data acquisition method, terminal and network side equipment combined with source channel coding

Abstract

The invention discloses a data acquisition method for joint source channel coding, a terminal and network side equipment, and belongs to the technical field of communication, and the data acquisition method for joint source channel coding in the embodiment of the invention comprises the steps that first equipment inputs first information into a first artificial intelligence AI unit to obtain second information; transmitting the second information to a second device by using a preset mode; wherein the first information comprises channel state information (CSI), the first AI unit is used for coding the first information, and the preset mode is used for adding channel influence.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of communication, and particularly relates to a data acquisition method of joint source channel coding, a terminal and a network side device. BACKGROUND

[0002] Channel State Information (CSI) feedback enhancement is a kind of bilateral model, and the training mode includes three kinds: unilateral training, joint training and separate training. The unilateral training is performed in one entity, and the joint training and the separate training belong to bilateral training, and information transmission is required between the network and the User Equipment (UE).

[0003] At present, Artificial Intelligence (AI) can be used to source encode the CSI to realize the CSI feedback enhancement. The AI-based CSI source encoding adopts a separate source channel coding transmission mode in the actual transmission process, that is, after source encoding by using AI, channel coding is used to increase the corresponding redundancy to resist the influence of the wireless channel. The data acquired in this way has high accuracy, which is not conducive to the training of the CSI feedback enhancement. How to acquire data that is conducive to the training of the CSI feedback enhancement has not been solved. SUMMARY

[0004] The embodiments of the application provide a data acquisition method of joint source channel coding, a terminal and a network side device, which can solve the problem of acquiring data that is conducive to the training of the CSI feedback enhancement.

[0005] In a first aspect, a data acquisition method of joint source channel coding is provided, comprising:

[0006] A first device inputs first information into a first Artificial Intelligence (AI) unit to obtain second information;

[0007] The second information is transmitted to a second device by using a preset mode;

[0008] The first information includes Channel State Information (CSI), the first AI unit is used for encoding the first information, and the preset mode is used for adding channel influence.

[0009] In a second aspect, a data acquisition method of joint source channel coding is provided, comprising:

[0010] A second device receives second information sent by a first device by using a preset mode;

[0011] The preset manner is used to add channel influence, the second information is obtained by the first device inputting the first information into the first AI unit, the first information includes CSI, and the first AI unit is used to encode the first information.

[0012] In a third aspect, a data acquisition apparatus for joint source-channel coding is provided, comprising:

[0013] A processing module is configured to obtain second information by inputting first information into a first artificial intelligence (AI) unit.

[0014] A transmission module is configured to transmit the second information to a second device using a preset manner.

[0015] The first information includes channel state information (CSI), the first AI unit is used to encode the first information, and the preset manner is used to add channel influence.

[0016] In a fourth aspect, a data acquisition apparatus for joint source-channel coding is provided, comprising:

[0017] A receiving module is configured to receive second information sent by a first device using a preset manner.

[0018] The preset manner is used to add channel influence, the second information is obtained by the first device inputting the first information into the first AI unit, the first information includes CSI, and the first AI unit is used to encode the first information.

[0019] In a fifth aspect, a data acquisition apparatus for joint source-channel coding is provided, and the apparatus is configured to perform the steps of the method according to the first aspect or the second aspect.

[0020] In a sixth aspect, a terminal is provided, comprising a processor and a memory, wherein the memory stores programs or instructions executable on the processor, and the programs or instructions are executed by the processor to implement the steps of the method according to the first aspect or the second aspect.

[0021] In a seventh aspect, a terminal is provided, comprising a processor and a communication interface, wherein the processor is configured to implement the steps of the method according to the first aspect or the second aspect, and the communication interface is configured to be coupled with the processor.

[0022] In an eighth aspect, a network-side device is provided, comprising a processor and a memory, wherein the memory stores programs or instructions executable on the processor, and the programs or instructions are executed by the processor to implement the steps of the method according to the first aspect or the second aspect.

[0023] In a ninth aspect, a network-side device is provided, including a processor and a communication interface, wherein the processor is configured to implement the steps of the method according to the first aspect or the second aspect, and the communication interface is configured to be coupled with the processor.

[0024] In a tenth aspect, a readable storage medium is provided, in which a program or instructions are stored, and the program or instructions are executed by a processor to implement the steps of the method according to the first aspect or the second aspect.

[0025] In an eleventh aspect, a wireless communication system is provided, including a first device and a second device, wherein the first device is configured to implement the steps of the method according to the first aspect, and the second device is configured to implement the steps of the method according to the second aspect.

[0026] In a twelfth aspect, a chip is provided, including a processor and a communication interface, wherein the communication interface is coupled with the processor, and the processor is configured to run a program or instructions to implement the method according to the first aspect or the second aspect.

[0027] In a thirteenth aspect, a computer program / program product is provided, which is stored in a storage medium, and is executed by at least one processor to implement the steps of the method according to the first aspect or the second aspect.

[0028] In the embodiments of the present application, the first device inputs first information into the first AI unit to obtain second information, and transmits the second information to the second device in a preset manner, wherein the first information includes CSI, the first AI unit is configured to encode the first information, and the preset manner is configured to add channel influence, so that the second device can obtain the second information with added channel influence and consider the influence of the channel in the AI-based encoding process, thereby realizing data acquisition of AI-based joint source channel coding and facilitating CSI feedback enhanced training. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 A block diagram of a wireless communication system to which the embodiments of the present application can be applied is shown;

[0030] Figure 2 A flowchart of a data acquisition method of joint source channel coding provided by the embodiments of the present application is shown;

[0031] Figure 3 Another flowchart of a data acquisition method of joint source channel coding provided by the embodiments of the present application is shown;

[0032] Figure 4 Another flowchart of a data acquisition method of joint source channel coding provided by the embodiments of the present application is shown;

[0033] Figure 5 This illustration shows another flowchart of the data acquisition method for joint source-channel coding provided in an embodiment of this application;

[0034] Figure 6 This illustration shows a schematic diagram of a data acquisition device with joint source-channel coding provided in an embodiment of this application;

[0035] Figure 7 This illustration shows another structural schematic diagram of the data acquisition device with joint source-channel coding provided in the embodiments of this application;

[0036] Figure 8 This illustration shows a structural diagram of a communication device provided in an embodiment of this application;

[0037] Figure 9 This illustration shows a hardware structure diagram of a terminal provided in an embodiment of this application;

[0038] Figure 10 This diagram illustrates the hardware structure of a network-side device according to an embodiment of this application. Detailed Implementation

[0039] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0040] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, not limited in number; for example, the first object can be one or more. Furthermore, "or" in this application indicates at least one of the connected objects. For example, the scope of protection for "A or B" covers at least three scenarios: Scenario 1: including A but not B; Scenario 2: including B but not A; Scenario 3: including both A and B. In addition, the terms "A and / or B," "at least one of A and B," and "at least one of A or B" also cover at least the above three scenarios. The character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0041] The term "instruction" in this application can be either a direct instruction (or explicit instruction) or an indirect instruction (or implicit instruction). A direct instruction can be understood as one in which the sender explicitly informs the receiver of specific information, the operation to be performed, or the requested result, etc., in the instruction sent. An indirect instruction can be understood as one in which the receiver determines the corresponding information based on the instruction sent by the sender, or makes a judgment and determines the operation to be performed or the requested result, etc., based on the judgment result.

[0042] It is worth noting that the technologies described in this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), or other systems. The terms "system" and "network" in this application are often used interchangeably, and the described technologies can be used with the systems and radio technologies mentioned above, as well as with other systems and radio technologies. The following description describes New Radio (NR) systems for illustrative purposes, and the term NR is used in most of the following description; however, these technologies can also be applied to systems other than NR systems, such as 6th generation (6G) radio systems. th Generation 6G communication system.

[0043] Figure 1This diagram illustrates a block diagram of a wireless communication system applicable to embodiments of this application. The wireless communication system includes a terminal 11 and a network-side device 12. The terminal 11 can be a mobile phone, tablet computer, laptop computer, notebook computer, personal digital assistant (PDA), handheld computer, netbook, ultra-mobile personal computer (UMPC), mobile internet device (MID), augmented reality (AR), virtual reality (VR) device, robot, wearable device, flight vehicle, vehicle user equipment (VUE), shipboard equipment, pedestrian user equipment (PUE), smart home devices (home appliances with wireless communication capabilities, such as refrigerators, televisions, washing machines, or furniture), game consoles, personal computers (PCs), ATMs, or self-service machines, etc. Wearable devices include: smartwatches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart chains, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, etc. Among these, in-vehicle devices can also be referred to as in-vehicle terminals, in-vehicle controllers, in-vehicle modules, in-vehicle components, in-vehicle chips, or in-vehicle units, etc. It should be noted that the specific type of terminal 11 is not limited in this application embodiment. Network-side equipment 12 may include access network equipment or core network equipment, wherein access network equipment may also be referred to as Radio Access Network (RAN) equipment, radio access network function, or radio access network unit. Access network equipment may include base stations, Wireless Local Area Network (WLAN) access points (AS), or Wireless Fidelity (WiFi) nodes, etc.The term "base station" can be referred to as Node B (NB), Evolved Node B (eNB), Next Generation Node B (gNB), New Radio Node B (NR Node B), Access Point, Relay Base Station (RBS), Serving Base Station (SBS), Base Transceiver Station (BTS), Radio Base Station, Radio Transceiver, Basic Service Set (BSS), Extended Service Set (ESS), Home Node B (HNB), Home Evolved Node B, Transmit / Receive Point (TRP), or any other suitable term in the relevant field, as long as the same technical effect is achieved. The term "base station" is not limited to any specific technical terminology. It should be noted that this application embodiment only uses a base station in an NR system as an example for description and does not limit the specific type of base station.

[0044] Core network equipment, also known as core network nodes, core network functions, or core network elements, includes, but is not limited to, at least one of the following: Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (or L-NEF), and Binding Support Function. Support Functions (BSF), Application Functions (AF), Location Management Functions (LMF), Gateway Mobile Location Centres (GMLC), and Network Data Analytics Functions (NWDAF), etc. It should be noted that this application embodiment only uses core network equipment in the NR system as an example and does not limit the specific type of core network equipment. If the name of the core network equipment mentioned in this application embodiment changes in subsequent protocol versions (e.g., 6G), it will still be within the scope of protection of this application.

[0045] Optionally, the core network equipment can be implemented by one or more functional modules in a single device, or by multiple devices working together; this application does not specifically limit this. It is understood that the aforementioned functional modules can be network elements in hardware devices, software functional modules running on dedicated hardware, or virtualized functional modules instantiated on a platform (e.g., a cloud platform).

[0046] The following description, in conjunction with the accompanying drawings, details the data acquisition method, terminal, and network-side equipment for joint source-channel coding provided in this application, through some embodiments and application scenarios.

[0047] In 3GPP Rel-18, CSI feedback enhancement has three training cooperation methods. The first method (Type 1) involves training within a single entity, similar to general AI algorithm training. After training, the trained models are deployed to the UE and network (NW) sides respectively via model transfer. The second method (Type 2) and the third method (Type 3) both require the transmission of encoded information and gradient information from the NW to the UE. However, since 3GPP Rel-18's CSI feedback enhancement uses AI for source coding, channel interference was not considered in the design of Type 2 and Type 3. Therefore, this application's embodiment performs AI coding on the CSI source and adds channel interference before transmission, realizing AI-based joint source-channel coding and solving the problem of not considering channel interference in the AI-based coding process in Type 2 and Type 3 scenarios.

[0048] Figure 2 This diagram illustrates a data acquisition method for joint source-channel coding provided in an embodiment of this application. This method 200 can be executed by a first device. Figure 2 As shown, the method may include the following steps.

[0049] S202: The first device inputs the first information into the first AI unit to obtain the second information.

[0050] The first device can be a terminal or a network-side device, without specific limitations. The first AI unit can be an encoder or a decoder, without specific limitations.

[0051] In this embodiment of the application, the first information may include CSI, the first AI unit is used to encode the first information, and the obtained second information includes the encoded CSI.

[0052] In one implementation, the first information may include, in addition to CSI, image information or voice information, etc., without limitation.

[0053] In this embodiment of the application, CSI may include at least one of the following parameters:

[0054] 1) Channel Quality Indicator (CQI);

[0055] 2) Precoding Matrix Indicator (PMI);

[0056] 3) CSI Reference Signal Resource Indicator (CSI-RS Resource Indicator, CRI);

[0057] 4) SSB Resource Indicator (SS / PBCH Block Resource Indicator, SSBRI);

[0058] 5) Layer Indicator (LI);

[0059] 6) Rank Indicator (RI);

[0060] 7) Layer 1 reference signal received power (L1-RSRP);

[0061] 8) Layer 1 signal to interference plus noise ratio (L1-SINR);

[0062] 9) Capability Index;

[0063] 10) Time-Difference Carrier-Phase (TDCP);

[0064] 11) Bare channel information;

[0065] 12) Transform domain channel information; for example, angle delay domain channel information;

[0066] 13) Processed raw channel information; for example, transforming the spatial frequency domain channel information to the angular time delay domain channel and / or truncation of the channel, etc.

[0067] 14) The source corresponding to PMI is the V matrix after Singular Value Decomposition (SVD).

[0068] In one implementation, any of the parameters included in CSI may be replaced by at least one of the normalized mean, variance, bias, and product factor of that parameter, without any specific limitation.

[0069] In this embodiment of the application, the second information may be contained in one of the following:

[0070] 1) Downlink Control Information (DCI);

[0071] 2) Medium Access Control Element (MAC CE);

[0072] 3) Radio Resource Control (RRC);

[0073] 4) Physical Uplink Shared Channel (PUSCH);

[0074] 5) Physical Downlink Shared Channel (PDSCH);

[0075] 6) Physical Downlink Control Channel (PDCCH);

[0076] 7) Physical Uplink Control Channel (PUCCH).

[0077] In this embodiment of the application, the second information can be configured by RRC or MAC CE to be sent periodically, semi-persistently, or aperiodically.

[0078] In the periodic transmission method, the transmission period is configured by RRC or MAC CE, and the second information is periodically transmitted according to the configured period.

[0079] In the semi-persistent transmission mode, after configuration, it is activated by MAC CE or DCI, and after activation, the second information is sent periodically until deactivation or after a specified period.

[0080] In the non-periodic transmission method, after configuration, it is activated by MAC CE or DCI and sends the second information a specified number of times. This specified number of times can be configured by the network, such as once or more, and is not specifically limited.

[0081] S204: Transmit the second information to the second device using a preset method.

[0082] The preset methods are used to add channel effects, and can include: lossless transmission through the channel, lossless transmission, and lossy transmission. Channel effects include at least the signal changes caused by channel transmission, such as at least one of the following: channel interference, channel fading, and channel noise.

[0083] In this embodiment of the application, step S204 can be implemented using a variety of transmission methods, including but not limited to one of the following transmission methods:

[0084] The first transmission method involves sending the second information to the second device using lossless transmission through the channel corresponding to the preset channel information. Lossless transmission means that the transmission is protected by channel coding and / or verification methods, allowing erroneous information to be retransmitted and ensuring error-free transmission. Channel coding protection can include increasing redundancy to reduce the coding rate; verification methods can include adding cyclic redundancy check (CRC). This scenario involves adding channel influence at the sending end, where the first device adds the corresponding channel influence to the second information based on the preset channel information before transmission. The advantage of this method is that when the actual channel used for transmission and the channel implementing joint source channel coding do not match, data acquisition can be performed using the preset channel.

[0085] Alternatively, the second information can be transmitted through the channel multiple times, and the information transmitted after each transmission can be sent to the second device using lossless transmission. In this case, the second device can obtain multiple pieces of second information with added channel influence, thus achieving the effect of data enhancement.

[0086] For example, if the second information is processed in five separate operations: it is sent to the second device via a channel and using lossless transmission, then the second device can receive five pieces of second information with added channel influence, thereby increasing the sample size of the data collection.

[0087] The second transmission method involves sending the second information to the second device using lossless transmission. Lossless transmission means that the transmission is protected by methods such as channel coding and / or checksum verification, allowing erroneous information to be retransmitted and ensuring error-free transmission. Channel coding protection can include increasing redundancy to reduce the code rate; checksum verification can include adding CRC (Common Regression Curve).

[0088] In this scenario, after receiving the second information, the second device can transmit it through the channel corresponding to the preset channel information to obtain the encoded information with added channel influence, thus completing data acquisition. This method involves the second device, i.e., the receiving end, transmitting the received second information through the appropriate channel, which is a method of adding channel influence at the receiving end. The advantage of this method is that when the actual transmission channel and the channel implementing joint source channel coding do not match, data acquisition can be performed through the preset channel. Furthermore, after completing data acquisition, the second device can also perform operations such as training the second AI unit, the specifics of which are not limited.

[0089] The third transmission method involves sending the second information to the second device using lossy transmission. Lossy transmission means that the transmission is not protected by channel coding and / or verification methods, and errors may occur during the transmission process. This lossy transmission method reduces the bandwidth used for data acquisition and also adds the influence of the channel characteristics actually used for transmission.

[0090] In this embodiment of the application, the lossless transmission method described above may include at least one of the following:

[0091] 1) Source coding. The second device then performs source decoding accordingly.

[0092] 2) Channel coding. The second device then performs channel decoding accordingly.

[0093] 3) Modulation. The second device then performs demodulation accordingly.

[0094] 4) Layer mapping. The second device performs layer mapping accordingly.

[0095] 5) Interleaving. The second device then performs deinterleaving accordingly.

[0096] 6) Filling. The second device then performs defilling accordingly.

[0097] 7) Drilling holes.

[0098] 8) Rate matching.

[0099] 9) Scrambling. The second device then performs descrambling accordingly.

[0100] 10) Precoding. The second device then performs receive merging accordingly.

[0101] The aforementioned lossless transmission method, through channel coding, can guarantee error-free transmission and improve coding quality.

[0102] In this embodiment of the application, the aforementioned lossy transmission method may include at least one of the following:

[0103] 1) Layer mapping. The second device then performs demapping accordingly.

[0104] 2) Interleaving. The second device then performs deinterleaving accordingly.

[0105] 3) Filling. The second device then performs defilling accordingly.

[0106] 4) Drilling holes.

[0107] 5) Rate matching.

[0108] 6) Scrambling. The second device then performs descrambling accordingly.

[0109] 7) Precoding. The second device then performs the receive-and-merge accordingly.

[0110] The lossy transmission method described above allows for training by adding channel influences to the actual channel environment used for transmission, when the channel characteristics are the same during the training and deployment phases. This reduces resource consumption during training and avoids the problem of being unable to perform backpropagation.

[0111] In this embodiment of the application, the first device can be a terminal or a network-side device, and the first information can be obtained using at least one of the following methods:

[0112] The first method of acquisition: Obtain the first information predefined by the protocol.

[0113] The second acquisition method: collecting first information. In this case, the first device collects the first information itself, so there is no need to obtain the first information from other devices. For example, the UE collects and obtains CSI information.

[0114] The third acquisition method: receiving first information sent by a second device. The second device can be a UE or an NW, the specific method is not limited. For example, the second device collects and obtains the first information and then sends it to the first device, enabling the first device to obtain the first information. For instance, the second device (NW) collects CSI information and sends it to the first device (UE), thereby allowing the UE to obtain the CSI information.

[0115] The fourth acquisition method: receiving the first information sent by the server of the first device. Here, the server of the first device refers to the server providing services to the first device, and both belong to the same network operator. For example, both the first device and its server are devices of network operator A.

[0116] The fifth method of acquisition: receiving the first information sent by a third-party server. Here, a third-party server refers to a server belonging to another network operator. For example, the first device is a device belonging to network operator A, and the third-party server is a device belonging to network operator B.

[0117] In the second acquisition method described above, since the first information is collected by the first device, in order for the second device to use the labels to train the second AI unit, the first device can also send the sample information collected each time to the second device. After receiving it, the second device uses the sample information as labels to train the second AI unit. It is worth mentioning that the first device can send the second information and the sample information together to the second device, or it can send the second information and the sample information to the second device separately; there is no specific limitation.

[0118] Besides the second acquisition method mentioned above, the first device does not need to send sample information to the second device. The second device can collect sample information itself or have it obtained from other devices; the specific method is not limited. It is worth noting that in this case, the first and second devices pre-agree on how to use the first information to avoid mismatch issues.

[0119] In this embodiment of the application, the second information may include at least one of a bit sequence and a symbol sequence. The bit sequence corresponds to bit-level joint source-channel coding, and the symbol sequence corresponds to symbol-level joint source-channel coding.

[0120] For example, the second information is either encoded CSI information containing a bit sequence or encoded CSI information containing a symbol sequence.

[0121] The method provided in this embodiment involves a first device inputting first information into a first AI unit to obtain second information, and then transmitting the second information to a second device using a preset method. The first information includes CSI (Content Sequence Information). The first AI unit encodes the first information, and the preset method adds channel influence, enabling the second device to obtain the second information with added channel influence. This allows the influence of the channel to be considered during the AI-based encoding process, realizing data acquisition for AI-based joint source-channel coding. This facilitates training with CSI feedback enhancement, improves the practicality of the trained model, and makes the model more widely applicable.

[0122] Figure 3 This diagram illustrates a data acquisition method for joint source-channel coding provided in an embodiment of this application. This method 300 can be executed by a first device. Figure 3 As shown, the method may include the following steps.

[0123] S302: The first device inputs the first information into the first AI unit to obtain the second information.

[0124] The first device can be a terminal or a network-side device, and is not specifically limited. The first information includes CSI, the first AI unit is used to encode the first information, and the second information includes the encoded CSI.

[0125] In this embodiment of the application, the first device obtaining the second information may be triggered by one of the following:

[0126] 1) Network instructions;

[0127] 2) Terminal reporting;

[0128] 3) Predefined events.

[0129] The predefined event that triggers the acquisition of the second information may include at least one of the following:

[0130] 1) In Wi-Fi mode;

[0131] 2) Battery level exceeds the second threshold;

[0132] 3) In charging state;

[0133] 4) Supports preset traffic patterns; for example, sending the second message does not incur traffic overhead.

[0134] 5) In preset flow mode;

[0135] 6) No business requirements; for example, no need to train the first AI unit or the second AI unit;

[0136] 7) Performance indicators are below the third threshold.

[0137] S304: Transmit the second information to the second device using a preset method.

[0138] The preset methods may include: lossless transmission via a channel, lossless transmission, and lossy transmission.

[0139] In this embodiment of the application, the above method may further include:

[0140] The first device transmits the first relevant information to the second device using either lossy or lossless transmission. This first relevant information is related to the first information and the encoding / decoding process. The first relevant information can also be used by the second device to train the second AI unit.

[0141] It is worth mentioning that the first device can send the second information and the first related information to the second device separately, or the first device can send the second information and the first related information to the second device together. This application embodiment does not specifically limit this.

[0142] In this embodiment of the application, the first relevant information may include at least one of the following:

[0143] 1) Samples from the CSI dataset;

[0144] 2) Sample IDs in the CSI dataset;

[0145] 3) Encoded dimensional information; for example, the encoded dimensional information is (16, 32, 32, 16), etc.

[0146] 4) Channel influence information;

[0147] 5) Rules for obtaining the first information.

[0148] In one implementation, the rules for acquiring the first information can be determined between the first device and the second device by means of instructions such as protocol, RRC, DCI, or Uplink Control Information (UCI).

[0149] In this embodiment, the rule for obtaining the first information may include an iteration count index, which indicates how many samples to obtain from the current position among samples arranged in a preset order. The current position is either the initial position or the position where the last sample acquisition ended. The iteration count refers to the number of training iterations performed by the first AI unit or the second AI unit. In the first training iteration, the current position is the initial position; in each subsequent training iteration, the current position is the position where the last sample acquisition ended. Additionally, the first device may also send the training iteration count to the second device.

[0150] For example, if the dataset containing the first information includes 10,000 samples arranged in a preset order, and it is agreed that 100 samples are retrieved each time in the order of arrangement, then the retrieval rule for the first information in the first related information can simply include the iteration number index. When the iteration number index is 1, the second device retrieves samples 1 to 100 from the dataset; when the iteration number index is 2, the second device retrieves samples 101 to 200 from the dataset; and so on, until the training is completed.

[0151] In this embodiment of the application, the acquisition of the first relevant information can be triggered by one of the following:

[0152] 1) Network instructions;

[0153] 2) Terminal reporting;

[0154] 3) Predefined events.

[0155] The predefined event that triggers the acquisition of the first relevant information may include at least one of the following:

[0156] 1) In Wi-Fi mode;

[0157] 2) Battery level exceeds the second threshold;

[0158] 3) In charging state;

[0159] 4) Supports preset traffic patterns; for example, sending the second message does not incur traffic overhead.

[0160] 5) In preset flow mode;

[0161] 6) No business requirements; for example, no need to train the first AI unit or the second AI unit;

[0162] 7) Performance indicators are below the third threshold.

[0163] In this embodiment of the application, a correspondence between the first relevant information and the second information can also be established. This correspondence can take various forms, including but not limited to the following:

[0164] 1) The correspondence between the first information and the encoded first information. For example, (CSI_1, encoded CSI_1), (CSI_2, encoded CSI_2), etc.

[0165] 2) The correspondence between the first information and the first information encoded via lossy transmission. For example, the first device uses the third transmission method mentioned above, that is, it uses lossy transmission to send the second information to the second device. Wherein, the channel corresponding to this lossy transmission is A, then the correspondence can be (CSI_1, CSI_1 encoded via channel A), (CSI_2, CSI_2 encoded via channel A), etc.

[0166] 3) The correspondence between the first information and the encoded first information via the channel corresponding to the preset channel information. For example, if the first device uses the first transmission method described above, that is, the second information is transmitted to the second device via the channel corresponding to the preset channel information using lossless transmission. Wherein, the channel corresponding to the preset channel information is B, then the correspondence can be (CSI_1, CSI_1 encoded via channel B), (CSI_2, CSI_2 encoded via channel B), etc.

[0167] 4) The correspondence between the first information, the encoded first information, and the added channel effects. For example, if the channel effect is specifically channel interference, the correspondence could be (CSI_2, encoded CSI_2, channel interference 2), etc.

[0168] It should be noted that in any of the four correspondence relationships mentioned above, the first information, such as CSI_1, can be replaced by the ID of the first information, such as CSI_ID_1. Accordingly, (CSI_1, encoded CSI_1) can be replaced by (CSI_ID_1, encoded CSI_1), and (CSI_1, encoded CSI_1, channel interference 1) can be replaced by (CSI_ID_1, encoded CSI_1, channel interference 1). This will not be explained in detail here.

[0169] This method of using the sample's ID instead of the sample for transmission eliminates the need to send the sample to a second device, thus reducing transmission overhead.

[0170] S306: Receive gradient information of the second information sent by the second device.

[0171] The gradient information of the second information is obtained by the second device training the second AI unit based on the second information.

[0172] In this embodiment, the second information and the gradient information of the second information are the information propagated forward and backward between the first device and the second device during training.

[0173] In the embodiments of this application, in the scenario of joint training, the first device can be a terminal and the second device can be a network-side device, or both the first device and the second device can be terminals. In the scenario of separate training, the first device can be a terminal and the second device can be a network-side device, or both the first device and the second device can be terminals, or the first device can be a network-side device and the second device can be a terminal.

[0174] In this embodiment, the second AI unit can be a decoder or an encoder, and the specific type is not limited. For example, in joint training or separate training scenarios, the second AI unit can be a decoder. In separate training scenarios, the second AI unit can be an encoder.

[0175] In this embodiment of the application, the gradient information of the second information can be carried in one of the following ways:

[0176] 1) DCI;

[0177] 2) MAC CE;

[0178] 3) RRC;

[0179] 4) PUSCH;

[0180] 5) PDSCH;

[0181] 6) PDCCH;

[0182] 7) PUCCH.

[0183] In this embodiment of the application, the gradient information of the second information can be configured by RRC or MAC CE to be periodically transmitted, semi-persistently transmitted, or non-periodically transmitted.

[0184] In the periodic transmission method, the transmission period is configured by RRC or MAC CE, and after configuration, the gradient information of the second information is periodically transmitted according to the period.

[0185] In the semi-persistent transmission mode, after configuration, it is activated by MAC CE or DCI, and after activation, it periodically transmits the gradient information of the second information until it is deactivated or after a specified period.

[0186] In the non-periodic transmission method, after configuration, it is activated by MAC CE or DCI and transmits the gradient information of the second information a specified number of times. This specified number of times can be configured by the network, such as once or more, and is not specifically limited.

[0187] In this embodiment, the gradient information of the second information can be greater than or equal to a preset first threshold. That is, before sending the gradient information of the second information, the second device filters and selects gradient information of the second information that is greater than or equal to the preset first threshold, and then sends it to the first device. Gradient information of the second information that is less than the preset first threshold is considered meaningless for training the first AI unit and therefore does not need to be sent to the first device.

[0188] S308: Train the first AI unit based on the gradient information of the second information.

[0189] In one implementation, step S308 may include: training a first AI unit based on the gradient information of the training parameter information and the second information. The training parameter information comprises various parameter information related to training the first AI unit.

[0190] In this embodiment of the application, the training parameter information may include at least one of the following:

[0191] 1) Batch dimensions;

[0192] 2) Optimizer type;

[0193] 3) Learning rate;

[0194] 4) Learning rate decay; for example, the exponential decay rate estimated for the first time, the exponential decay rate estimated for the second time, the exponential decay rate estimated for the Nth time, etc.

[0195] 5) Loss function; where the loss function may include a combination function of at least one of the following: the error between the target model's output and the label, mean squared error, normalized mean squared error, correlation, entropy, mutual information, and constant. The target model is either the first AI unit or the second AI unit. This combination function includes at least one of the following: a combination of various common mathematical operations such as addition, subtraction, multiplication, division, Nth power, Nth square root, logarithm, differentiation, and partial differentiation. N is any number, such as positive, negative, or 0, real or complex.

[0196] 6) Regularization methods and parameters;

[0197] 7) Initialization methods and parameters;

[0198] 8) Dropout methods and parameters;

[0199] 9) Training cycle;

[0200] 10) Number of iterations;

[0201] 11) Training termination conditions.

[0202] In one implementation, the loss function can be obtained by weighted combination of loss information from multiple parts of the target model's output. This weighted combination can be a combination of linear averaging, multiplicative averaging, and other common averaging methods. The multiple parts are obtained by partitioning according to at least one of spatial domain resources, code domain resources, frequency domain resources, and time domain resources. Frequency domain resources can include resource blocks (RBs), subbands, or precoding resource block groups (PRGs). Time domain resources can include subcarriers, symbols, time slots, or half-time slots. The loss information can include loss values ​​and / or loss-related functions. The target model is a first AI unit or a second AI unit.

[0203] In this embodiment of the application, the training parameter information may be preset by the protocol, or indicated by the first device or the second device using at least one of the following methods:

[0204] 1) RRC indication;

[0205] 2) MAC CE indication;

[0206] 3) Control information indication; for example, DCI indication of NW or UCI indication of UE.

[0207] S310: Establish the matching relationship between the first and second AI units after training.

[0208] In this embodiment of the application, the identifier of the matching relationship may include at least one of the following:

[0209] 1) Model ID;

[0210] 2) Functionality ID;

[0211] 3) Dataset ID;

[0212] 4) Pairing ID;

[0213] 5) Associate ID.

[0214] In one implementation, the second information may include a bit sequence. Before transmitting the second information to the second device using a lossy transmission method, the method further includes: modulating the bit sequence in the second information using at least one of the following modulation methods:

[0215] 1) Binary Phase Shift Keying (BPSK);

[0216] 2)pi / 2-BPSK;

[0217] 3) Quadrature Phase Shift Keying (QPSK);

[0218] 4) Quadrature Amplitude Modulation (QAM);

[0219] 5) 64QAM;

[0220] 6) 256QAM;

[0221] 7) 1024QAM;

[0222] 8)4096QAM;

[0223] 9) AI-based modulation.

[0224] In this embodiment of the application, the above method may further include:

[0225] The first device transmits preset channel information to the second device using either lossy or lossless transmission. The preset channel information is used to add channel effects to the second information using the corresponding channel.

[0226] In this embodiment of the application, the preset channel information may include at least one of the following:

[0227] 1) Signal-to-noise ratio (SNR);

[0228] 2) SNR range;

[0229] 3) Bit error rate;

[0230] 4) Bit error rate range;

[0231] 5) CQI;

[0232] 6) CQI range;

[0233] 7) Modulation and Coding Scheme (MCS);

[0234] 8) MCS range;

[0235] 9) Raw channel information of the transmission channel; wherein, the raw channel information of the transmission channel can be set in the form of a dataset and associated with an ID;

[0236] 10) Channel model.

[0237] In one implementation, the preset channel information can be preset by the protocol. For example, the protocol may specify that the preset channel used by the first device and the second device is an Additive White Gaussian Noise (AWGN) channel.

[0238] In another implementation, the preset channel information can be indicated by the first device or the second device using at least one of the following methods:

[0239] 1) Radio Resource Control (RRC) indication;

[0240] 2) Medium Access Control Element (MAC CE) indication;

[0241] 3) Control information indication; for example, DCI indication of NW or UCI indication of UE.

[0242] In this embodiment of the application, the training of the first AI unit or the training of the second AI unit can be triggered by one of the following:

[0243] 1) Network instructions;

[0244] 2) Terminal reporting;

[0245] 3) Predefined events.

[0246] The predefined event that triggers the training of the first AI unit or the second AI unit may include at least one of the following:

[0247] 1) In Wi-Fi mode;

[0248] 2) Battery level exceeds the second threshold;

[0249] 3) In charging state;

[0250] 4) Supports preset traffic patterns; for example, sending the second message does not incur traffic overhead.

[0251] 5) In preset flow mode;

[0252] 6) No business requirements; for example, no need to train the first AI unit or the second AI unit;

[0253] 7) Performance indicators are below the third threshold.

[0254] The method provided in this application embodiment involves inputting first information into a first AI unit via a first device to obtain second information, transmitting the second information to a second device using a preset method, receiving gradient information of the second information sent by the second device, training the first AI unit based on training parameter information and the gradient information of the second information, and establishing a matching relationship between the trained first AI unit and the second AI unit. The first information is CSI (Content Sequence Information), the first AI unit is used to encode the first information, and the preset method is used to add channel influence, enabling the second device to obtain the second information with added channel influence. This allows the influence of the channel to be considered during AI-based encoding, realizing data acquisition for AI-based joint source-channel coding, facilitating CSI feedback enhancement training, improving the practicality of the trained model, and making the model more widely applicable.

[0255] Figure 4 This diagram illustrates a data acquisition method for joint source-channel coding provided in an embodiment of this application. This method 400 can be executed by a second device. Figure 4 As shown, the method may include the following steps.

[0256] S402: The second device receives the second information sent by the first device using a preset method.

[0257] The second device can be a terminal or a network-side device, with no specific limitation. Preset methods are used to add channel impact, including: lossless transmission through the channel, lossless transmission, and lossy transmission. Channel impact includes at least signal changes caused by channel transmission, such as channel interference, channel fading, and channel noise.

[0258] The second information is obtained by the first device inputting the first information into the first AI unit. The first information includes CSI. The first AI unit is used to encode the first information. The second information includes the encoded CSI.

[0259] In this embodiment, the second information may include at least one of a bit sequence and a symbol sequence. The bit sequence corresponds to bit-level joint source-channel coding, and the symbol sequence corresponds to symbol-level joint source-channel coding. The second information may be carried in one of the following: DCI, MAC CE, RRC, PUSCH, PDSCH, PDCCH, or PUCCH. The second information may be configured by RRC or MAC CE to be transmitted periodically, semi-persistently, or aperiodically.

[0260] In one implementation, the second information can be used by the second device to train the second AI unit.

[0261] In this embodiment of the application, step S402 may include one of the following:

[0262] The second device receives the second information sent by the first device through the corresponding channel according to the preset channel information and using lossless transmission.

[0263] The second device receives the second information sent by the first device using lossless transmission, and transmits the second information through the corresponding channel according to the preset channel information.

[0264] The second device receives the second information sent by the first device using a lossy transmission method.

[0265] In one implementation, after the second information is transmitted through the corresponding channel according to the preset channel information, the second device can also perform the operation of training the second AI unit.

[0266] In this embodiment of the application, the above-mentioned preset method is a lossy transmission method. Accordingly, after step S402, at least one of the following operations is performed on the second information:

[0267] 1) Demapping; wherein, the first device performs layer mapping during lossy transmission;

[0268] 2) Deinterleaving; wherein, the first device performs interleaving during lossy transmission;

[0269] 3) De-padded; wherein, the first device performs padding during lossy transmission;

[0270] 4) Descrambling; wherein, the first device performs scrambling during lossy transmission;

[0271] 5) Receive and merge; wherein, the first device performs precoding during lossy transmission.

[0272] In this embodiment of the application, the preset method mentioned above includes a lossless transmission method. Accordingly, after step S402, at least one of the following operations is performed on the second information:

[0273] 1) Source decoding; wherein, the first device performs source encoding during lossless transmission;

[0274] 2) Channel decoding; wherein, the first device performs channel coding during lossless transmission;

[0275] 3) Demodulation; wherein, the first device performs modulation during lossless transmission;

[0276] 4) Demapping; wherein, the first device performs layer mapping during lossless transmission;

[0277] 5) Deinterleaving; wherein, the first device performs interleaving during lossless transmission;

[0278] 6) De-filling; wherein, the first device performs filling during lossless transmission;

[0279] 7) Descrambling; wherein, the first device performs scrambling during lossless transmission;

[0280] 8) Receive and merge; wherein, the first device performs precoding during lossless transmission.

[0281] The method provided in this application embodiment receives second information sent by a first device using a preset method through a second device. The preset method is used to add channel influence. The second information is obtained by the first device by inputting the first information into a first AI unit. The first information is CSI. The first AI unit is used to encode the first information so that the second device can obtain the second information with added channel influence. It can consider the influence of the channel in the AI-based encoding process, realize the data acquisition of AI-based joint source-channel coding, help the training of CSI feedback enhancement, improve the practicality of the training model, and make the model more widely used.

[0282] Figure 5 This diagram illustrates a data acquisition method 500 for joint source-channel coding provided in an embodiment of this application. This method 500 can be executed by a second device. Figure 5 As shown, the method may include the following steps.

[0283] S502: The second device receives the second information sent by the first device using a preset method.

[0284] The second device can be a terminal or a network-side device, with no specific limitation. Preset methods are used to add channel impact, including: lossless transmission through the channel, lossless transmission, and lossy transmission. Channel impact includes at least signal changes caused by channel transmission, such as channel interference, channel fading, and channel noise.

[0285] The second information is obtained by the first device inputting the first information into the first AI unit. The first information includes CSI. The first AI unit is used to encode the first information. The second information includes the encoded CSI.

[0286] S504: Input the second information into the second AI unit for decoding to obtain the recovered first information.

[0287] S506: Based on the training parameter information, the first relevant information, and the recovered first information, train the second AI unit to obtain the gradient information of the second information.

[0288] The training parameter information refers to the parameter information related to the training of the second AI unit. The training parameter information may include at least one of the following: batch size, optimizer type, learning rate, learning rate decay, loss function, regularization method and parameters, initialization method and parameters, dropout method and parameters, training period, number of iterations, and training termination condition.

[0289] In one implementation, the loss function can be obtained by weighted combination of loss information from multiple parts of the target model's output. This weighted combination can be a combination of linear averaging, multiplicative averaging, and other common averaging methods. The multiple parts are obtained by partitioning according to at least one of spatial domain resources, code domain resources, frequency domain resources, and time domain resources. Frequency domain resources can include resource blocks (RBs), subbands, or precoding resource block groups (PRGs). Time domain resources can include subcarriers, symbols, time slots, or half-time slots. The loss information can include loss values ​​and / or loss-related functions. The target model is a first AI unit or a second AI unit.

[0290] In this embodiment of the application, the training parameter information can be preset by the protocol, or indicated by the first device or the second device using at least one of the following methods: RRC indication, MAC CE indication, and control information indication. The control information indication includes: NW's DCI indication or UE's UCI indication.

[0291] In this embodiment of the application, the above method may further include: the second device receiving first relevant information sent by the first device using lossy transmission or lossless transmission.

[0292] The first relevant information refers to information related to the first information and the encoding / decoding. The first relevant information may include at least one of the following: samples in the CSI dataset, sample IDs in the CSI dataset, encoded dimensional information, channel influence information, and the rules for obtaining the first information.

[0293] In one implementation, the rules for acquiring the first information can be determined between the first device and the second device through instructions such as protocols, RRC, DCI, or UCI.

[0294] In this embodiment, the rule for obtaining the first information may include an iteration count index, which indicates how many samples to obtain from the current position among samples arranged in a preset order. The current position is either the initial position or the position where the last sample acquisition ended. The iteration count refers to the number of training iterations performed by the first AI unit or the second AI unit. In the first training iteration, the current position is the initial position; in each subsequent training iteration, the current position is the position where the last sample acquisition ended. Additionally, the first device may also send the training iteration count to the second device.

[0295] In this embodiment, the acquisition of the first relevant information can be triggered by one of the following: network indication, terminal reporting, or predefined event. The predefined event may include at least one of the following: being in Wi-Fi mode, battery level exceeding a second threshold, being in charging mode, supporting a preset data traffic mode, being in a preset data traffic mode, having no service requirements, or performance indicators falling below a third threshold.

[0296] S508: Send the gradient information of the second information to the first device.

[0297] The gradient information of the second information is used by the first device to train the first AI unit in combination with the training parameter information.

[0298] In this embodiment, the gradient information of the second information can be carried in one of the following: DCI, MAC CE, RRC, PUSCH, PDSCH, PDCCH, or PUCCH. The gradient information of the second information can be configured by RRC or MAC CE to be transmitted periodically, semi-persistently, or aperiodically.

[0299] In this embodiment, the gradient information of the second information can be greater than or equal to a preset first threshold. That is, before sending the gradient information of the second information, the second device filters and selects gradient information of the second information that is greater than or equal to the preset first threshold, and then sends it to the first device. Gradient information of the second information that is less than the preset first threshold is considered meaningless for training the first AI unit and therefore does not need to be sent to the first device.

[0300] In this embodiment of the application, the above method may further include:

[0301] The second device receives preset channel information sent by the first device using lossy or lossless transmission methods.

[0302] Among them, the preset channel information is used to add channel influence to the second information using the corresponding channel.

[0303] In this embodiment of the application, the preset channel information may include at least one of the following: SNR, SNR range, bit error rate, bit error rate range, CQI, CQI range, MCS, MCS range, raw channel information of the transmission channel, and channel model.

[0304] In this embodiment of the application, the preset channel information can be indicated by the first device or the second device using at least one of the following methods: protocol indication, RRC indication, MAC CE indication, DCI indication, and UCI indication.

[0305] The method provided in this application embodiment involves a second device receiving second information sent by a first device using a preset method, inputting the second information into a second AI unit for decoding to obtain recovered first information, training the second AI unit based on training parameter information, first related information, and the recovered first information to obtain gradient information of the second information, and sending the gradient information of the second information to the first device. The preset method is used to add channel influence, the second information is obtained by the first device inputting the first information into the first AI unit, the first information being CSI, and the first AI unit encoding the first information. This enables the second device to obtain second information with added channel influence, allowing the channel influence to be considered during AI-based encoding. This achieves data acquisition for AI-based joint source-channel coding, facilitates CSI feedback enhancement training, improves the practicality of the trained model, and makes the model more widely applicable.

[0306] The following description addresses the AI ​​units and AI training involved in the above embodiments.

[0307] In this application embodiment, AI can also refer to Machine Learning (ML). An AI unit (including a first AI unit and a second AI unit) can also be called an AI model, AI structure, etc., or the AI ​​unit can refer to a processing unit capable of implementing specific algorithms, formulas, processing flows, capabilities, etc. related to AI, or the AI ​​unit can be a processing method, algorithm, function, module, or unit for a specific dataset, or the AI ​​unit can be a processing method, algorithm, function, module, or unit running on AI-related hardware such as GPUs, NPUs, TPUs, and ASICs. This application does not specifically limit this. Optionally, the specific dataset includes the input and / or output of the AI ​​unit.

[0308] Optionally, the identifier of the AI ​​unit may be one or more of the following identifiers: AI unit identifier, AI structure identifier, AI algorithm identifier, function identifier, physical identifier, logical identifier, global identifier, local identifier, identifier of the specific dataset associated with the AI ​​unit, identifier of the specific scenario, environment, channel characteristics, or device related to the AI, or identifier of the function, characteristic, capability, or module related to the AI. This application does not specifically limit these identifiers.

[0309] In this embodiment, the first AI unit and the second AI unit can align third information before training to ensure training accuracy. The third information may include at least one of the following: structural features of the model, the model's load quantization method, the model's estimation accuracy, and the model's output accuracy. Here, "model" refers to either the first AI unit or the second AI unit.

[0310] In this embodiment of the application, the structural features of the above model may include at least one of the following: model structure, basic structural features of the model, structural features of model sub-modules, number of model layers, number of model neurons, model size, model complexity, and quantization parameters of model parameters.

[0311] In this embodiment of the application, the basic structural features of the above-mentioned model may include at least one of the following: whether it includes a fully connected structure, whether it includes a convolutional structure, whether it includes a long short-term memory (LSTM) structure, whether it includes an attention structure, and whether it includes a residual structure.

[0312] In this embodiment, the number of neurons in the model may include at least one of the following: the number of fully connected neurons, the number of convolutional neurons, the number of memory neurons, the number of attention neurons, and the number of residual neurons. And / or, the number of neurons in the model may include at least one of the following: the number of neurons of all types, the number of neurons of a single type, the total number of neurons in the entire model, and the number of neurons in a single layer or several layers.

[0313] In this embodiment of the application, the quantization parameters of the above-mentioned model parameters may include at least one of the following: the quantization method of the model parameters and the number of quantization bits of a single neuron parameter.

[0314] The quantization method of the model parameters may include at least one of the following: uniform quantization, non-uniform quantization, weight-sharing quantization or group quantization, parameter encoding quantization, transform domain quantization, and product quantization.

[0315] In the embodiments of this application, the above-mentioned load quantization method may include at least one of the following: quantization method, dimension of features before and after quantization, and quantization method used during quantization.

[0316] The quantization method used in the above quantization may include at least one of the following: when using codebook for quantization, the codebook content and codebook usage method need to be synchronized; when using specific rules for quantization, the quantization rules need to be synchronized.

[0317] The quantization rules mentioned above may include at least one of the following: N quantization intervals and a quantization method, where N is a positive integer. The quantization method may include at least one of the following: uniform quantization, non-uniform quantization, weight-sharing quantization or grouped quantization, parameter-encoded quantization, transform-domain quantization, or product quantization.

[0318] In the context of synchronizing codebook content and codebook usage methods, and / or synchronizing quantization rules, the synchronization method may include any of the following: selecting a set number representing the selected method from a predefined set of methods, or directly sending the codebook content.

[0319] In this embodiment of the application, the method by which the first device and the second device align the third information may include at least one of the following:

[0320] When the first device or other network element sends the target information to the second device, it also sends the third information.

[0321] When the second device or other network element sends the target information to the first device, it also sends the third information.

[0322] Before the first device or other network element sends the target information to the second device, the first device or other network element sends the third information.

[0323] Before the second device or other network element sends the target information to the first device, the second device or other network element sends the third information.

[0324] When requesting target information, the second device sends third information;

[0325] When requesting target information, the first device sends third information;

[0326] When the second device requests target information, the first device or other network element sends consent information and sends third information. The consent information is used to indicate consent to the second device's request.

[0327] When the first device requests target information, the second device or other network element sends consent information and sends third information. The consent information is used to indicate consent to the first device's request.

[0328] The target information includes at least one first piece of information related to the first action of the model and at least one second piece of information corresponding to the at least one first piece of information.

[0329] In this embodiment of the application, after the first device sends confirmation information for the third information, the second device or other network element sends the target information; and / or, after the second device sends confirmation information for the third information, the first device or other network element sends the target information.

[0330] In this embodiment of the application, the method by which the first device and the second device align the third information includes at least one of the following:

[0331] After a device that has received the third information sends a confirmation message for the third information, the first AI unit and / or the second AI unit can use the model associated with the third information;

[0332] After a device that has received the third information sends a confirmation message for the third information and after a first period of time has elapsed, the first AI unit and / or the second AI unit can use the model associated with the third information.

[0333] After a first duration has elapsed since the third information was sent or received, the first AI unit and / or the second AI unit can use the model associated with the third information.

[0334] The first duration can be determined by any of the following: carried by third information, carried by confirmation information of third information, carried by other related information or signaling of third information, as agreed upon in the agreement, or determined by the capabilities of the first or second device.

[0335] In this embodiment of the application, joint source-channel coding can be defined or described in the following ways:

[0336] A) Inputs and outputs of the AI / ML unit;

[0337] B) A signal processing procedure;

[0338] C) Input / output information mapping.

[0339] The inputs and outputs of the AI / ML unit may include the following:

[0340] A1) AI / ML encoding methods, specifically including:

[0341] a. AI / ML encoding unit;

[0342] b. AI / ML decoding unit.

[0343] A2) The input to the AI / ML coding unit is CSI source information, which may include at least one of the following:

[0344] a.CQI;

[0345] b. PMI;

[0346] c.CRI;

[0347] d.SSBRI;

[0348] e.LI;

[0349] f.RI;

[0350] g.L1-RSRP;

[0351] h.L1-SINR;

[0352] i. Capability Index;

[0353] j.TDCP;

[0354] k. Bare channel information;

[0355] l. Processed raw channel information, such as transforming spatial frequency domain channel information to angular time delay domain channel and / or truncating the channel;

[0356] The information source corresponding to m.PMI is the V matrix after the channel SVD decomposition.

[0357] A3) The output of the AI / ML coding unit is the bit sequence after joint source-channel coding (bit-level joint source-channel coding); or,

[0358] The output of the AI / ML coding unit is the symbol sequence after joint source-channel coding (symbol-level joint source-channel coding).

[0359] A4) The output of the AI / ML coding unit cannot use channel coding.

[0360] The input to the A5) AI / ML decoding unit includes one of the following:

[0361] a. The bit sequence received by NW that is related to the output of the AI / ML encoder;

[0362] b. The symbol sequence received by NW that is associated with the AI / ML encoder output.

[0363] A6) The output of the AI / ML decoding unit is the recovered CSI source information, which may include at least one of the following:

[0364] a.CQI;

[0365] b. PMI;

[0366] c.CRI;

[0367] d.SSBRI;

[0368] e.LI;

[0369] f.RI;

[0370] g.L1-RSRP;

[0371] h.L1-SINR;

[0372] i. Capability Index;

[0373] j.TDCP;

[0374] k. Bare channel information;

[0375] l. Processed raw channel information, such as transforming spatial frequency domain channel information to angular time delay domain channel and / or truncating the channel;

[0376] The information source corresponding to m.PMI is the V matrix after the channel SVD decomposition.

[0377] The aforementioned signal processing flow may include the following four components:

[0378] B1) A joint source-channel coding process for CSI information feedback, comprising the following modules:

[0379] a. Joint source channel coding processing module, located on the UE side;

[0380] b. Joint source channel decoding and processing module, located on the NW side, such as on the gNB side.

[0381] B2) Perform channel estimation on CSI-RS to obtain CSI information, which may include at least one of the following:

[0382] a.CQI;

[0383] b. PMI;

[0384] c.CRI;

[0385] d.SSBRI;

[0386] e.LI;

[0387] f.RI;

[0388] g.L1-RSRP;

[0389] h.L1-SINR;

[0390] i. Capability Index;

[0391] j.TDCP;

[0392] k. Bare channel information;

[0393] l. Processed raw channel information, such as transforming spatial frequency domain channel information to angular time delay domain channel and / or truncating the channel;

[0394] The information source corresponding to m.PMI is the V matrix after the channel SVD decomposition.

[0395] B3) The UE side uses the joint source channel coding processing module to process CSI information to obtain coded information, which may include one of the following:

[0396] a. Bit sequence (bit-level joint source coding);

[0397] b. Symbol sequence (symbol-level joint source coding).

[0398] B4) The NW side uses a joint source-channel decoding processing module to process the coded bit sequence or symbol sequence containing channel interference and noise, which can obtain the recovered CSI information.

[0399] The aforementioned input / output information mapping may include one of the following:

[0400] C1) A joint source-channel coding mapping on the UE side, which maps CSI information to a sequence of information bits after joint source-channel coding (bit-level joint source-channel coding) or a sequence of symbols (symbol-level joint source-channel coding).

[0401] C2) A joint source-channel decoding mapping on the NW side maps a bit sequence (bit-level joint source-channel coding) or symbol sequence (symbol-level joint source-channel coding) containing channel interference and noise into recovered CSI information.

[0402] The application of the above embodiments is explained in detail below with reference to six scenarios.

[0403] The embodiments of this application involve the following two encoding methods: Joint Source Channel Coding (JSCC) and Separate Source Channel Coding (SSCC).

[0404] In this embodiment of the application, the first device transmits the second information to the second device using a preset method, including but not limited to one of the following three transmission methods:

[0405] The first transmission method (the method of adding channel influence at the sending end): After the second information is transmitted through the channel corresponding to the preset channel information, it is sent to the second device using a lossless transmission method.

[0406] The second transmission method (the method of adding channel influence at the receiving end): The second information is sent to the second device using lossless transmission. In this method, after receiving the second information, the second device can transmit it through the channel corresponding to the preset channel information to obtain the encoded information with added channel influence, thus completing data acquisition.

[0407] The third transmission method (lossy transmission method): The second information is sent to the second device using a lossy transmission method.

[0408] Combining the two encoding methods and three transmission methods mentioned above, we can derive a variety of scenarios. The following are six examples illustrating these scenarios.

[0409] Scenario 1: Symbol-level JSCC joint training, using a lossy transmission method;

[0410] Scenario 2: Bit-level JSCC joint training, where the transmitting UE adds channel influence.

[0411] Scenario 3: Symbol-level JSCC joint training, where the receiver NW adds channel influence;

[0412] Scenario 4: Symbol-level JSCC Separate training, training the decoder on the receiver NW side, and adding channel effects on the NW side;

[0413] Scenario 5: Separate training of symbol-level JSCC, training the decoder on the receiver NW side, and adding channel influence on the transmitter UE side;

[0414] Scenario 6: Symbol-level JSCC Separate training, a method of training the encoder on the UE side without adding channel influence.

[0415] In Scenario 1, the first device is the UE (User Equipment) and the second device is the NW (Network Controller). The channels of the UE and NW are approximately the same as those in actual deployment during joint training. Furthermore, the UE and NW are familiar with the selected training dataset, and symbol-level JSCC Joint Training is used. The data acquisition method for joint source-channel coding in this scenario can include the following process:

[0416] Step 1: The UE selects samples with a batch size of 128 from the CSI dataset, records the sample ID (e.g., 3, 6, ...), and obtains the first information, namely the CSI information to be encoded.

[0417] Step 2: The UE inputs the CSI information to be encoded into the first AI unit, i.e., the symbol-level joint source channel encoder, to obtain the second information, i.e., the encoded CSI symbol information.

[0418] Step 3: The UE performs layer / resource mapping on the encoded CSI symbol information and sends it to the NW via lossy transmission.

[0419] Step 4: The UE sends the first relevant information to the NW through lossless transmission. The first relevant information includes the shape and sample ID of the encoded CSI symbol information.

[0420] Step 5: The NW receives the encoded CSI symbol information and the first related information. After demapping the encoded CSI symbol information, it uses the first related information to perform a reshape operation on the encoded CSI symbol information to obtain the original shape. Then, it inputs the original shape into the second AI unit, i.e., the symbol-level joint source-channel decoder, to obtain the recovered CSI information.

[0421] Step 6: NW obtains the CSI information to be encoded based on the sample ID. Based on the training parameter information, it uses the loss function to calculate the loss between the recovered CSI information and the CSI information to be encoded. Then, it uses the backpropagation algorithm to train the symbol-level joint source-channel decoder to obtain the gradient information of the encoded CSI symbol information.

[0422] Step 7: The NW sends the gradient information of the encoded CSI symbol information to the UE through lossless transmission.

[0423] Step 8: The UE receives the gradient information of the encoded CSI symbol information and trains the symbol-level joint source-channel encoder based on the training parameter information and the gradient information of the encoded CSI symbol information.

[0424] Step 9: Repeat steps 1 to 8 above until the training termination condition is met or the training iterations are completed.

[0425] Step 10: Establish the matching relationship between the trained UE-side symbol-level joint source-channel encoder and the NW-side symbol-level joint source-channel decoder.

[0426] In Scenario 2: the first device is the UE, and the second device is the NW. A bit-level JSCC Joint training method is used, adding channel influence at the UE side of the transmitter. The NW is unaware of the training dataset selected by the UE. The data acquisition method for joint source-channel coding in this scenario can include the following process:

[0427] Step 1: The UE selects a batch of 128 samples from the CSI dataset to obtain the first information, namely the CSI information to be encoded.

[0428] Step 2: The UE inputs the CSI information to be encoded into the first AI unit, i.e., the bit-level joint source channel encoder, to obtain the second information, i.e., the encoded CSI bit information.

[0429] Step 3: The UE transmits the encoded CSI bit information through the corresponding channel according to the first channel information, adds channel effects, obtains the encoded CSI bit information transmitted through the channel, and then sends it to the NW through lossless transmission.

[0430] Step 4: The UE sends the first relevant information to the NW through lossless transmission. The first relevant information includes the shape of the encoded CSI bit information and the CSI sample information.

[0431] Step 5: The NW receives the encoded CSI bit information and the first related information passed through the channel. It uses the first related information to perform a reshape operation on the encoded CSI bit information passed through the channel. After obtaining the original shape, it inputs it into the second AI unit, i.e., the bit-level joint source-channel decoder, to obtain the recovered CSI information.

[0432] Step 6: Based on the training parameter information, NW uses the loss function to calculate the loss between the recovered CSI information and the CSI sample information, and uses the backpropagation algorithm to train the bit-level joint source-channel decoder to obtain the gradient information of the encoded CSI bit information.

[0433] Step 7: The NW sends the gradient information of the encoded CSI bit information to the UE through lossless transmission.

[0434] Step 8: The UE receives the gradient information of the encoded CSI bit information and trains a bit-level joint source-channel encoder based on the training parameter information and the gradient information of the encoded CSI bit information.

[0435] Step 9: Repeat steps 1 to 8 above until the training termination condition is met or the training iterations are completed.

[0436] Step 10: Establish the matching relationship between the trained UE-side bit-level joint source-channel encoder and the NW-side bit-level joint source-channel decoder.

[0437] In Scenario 3, the first device is the UE (User Equipment), and the second device is the NW (Network Controller). A symbol-level JSCC (Joint Source-Channel Coding) training method is used, adding channel influences at the receiving NW side. The NW is unaware of the training dataset selected by the UE. The data acquisition method for joint source-channel coding in this scenario can include the following process:

[0438] Step 1: The UE selects a batch of 128 samples from the CSI dataset to obtain the first information, namely the CSI information to be encoded.

[0439] Step 2: The UE inputs the CSI information to be encoded into the first AI unit, i.e., the symbol-level joint source channel encoder, to obtain the second information, i.e., the encoded CSI symbol information.

[0440] Step 3: The UE sends the encoded CSI symbol information to the NW through lossless transmission. After receiving it, the NW transmits the encoded CSI symbol information through the corresponding channel according to the first channel information.

[0441] Step 4: The UE sends the first relevant information to the NW through lossless transmission. The first relevant information includes the shape of the encoded CSI symbol information and the CSI sample information.

[0442] Step 5: The NW uses the first relevant information to perform a reshape operation on the encoded CSI symbol information passing through the channel. After obtaining the original shape, it is input into the second AI unit, i.e., the symbol-level joint source-channel decoder, to obtain the recovered CSI information.

[0443] Step 6: Based on the training parameter information, NW uses the loss function to calculate the loss between the recovered CSI information and the CSI information to be encoded, and uses the backpropagation algorithm to train the symbol-level joint source-channel decoder to obtain the gradient information of the encoded CSI symbol information.

[0444] Step 7: The NW sends the gradient information of the encoded CSI symbol information to the UE through lossless transmission.

[0445] Step 8: The UE receives the gradient information of the encoded CSI symbol information and trains the symbol-level joint source-channel encoder based on the training parameter information and the gradient information of the encoded CSI symbol information.

[0446] Step 9: Repeat steps 1 to 8 above until the training termination condition is met or the training iterations are completed.

[0447] Step 10: Establish the matching relationship between the trained UE-side symbol-level joint source-channel encoder and the NW-side symbol-level joint source-channel decoder.

[0448] In Scenario 4: the first device is the UE, and the second device is the NW. Symbol-level SSCC Separate training is used. Before separate training, the UE has already trained its symbol-level joint source-channel encoder / decoder using the CSI dataset and preset channel information, eliminating the need for further encoder training. Furthermore, channel effects are added at the receiving NW side, which then trains the decoder. The data acquisition method for joint source-channel coding in this scenario can include the following process:

[0449] Step 1: The UE uses the first AI unit, i.e., the symbol-level joint source channel encoder, to encode all samples in the first information, i.e., the CSI dataset, to obtain the second information, i.e., the encoded CSI dataset.

[0450] Step 2: The UE sends the encoded CSI dataset to the NW through lossless transmission. The NW then transmits the encoded CSI dataset through the corresponding channel according to the first channel information to obtain the encoded CSI dataset transmitted through the channel.

[0451] Step 3: The UE sends the first relevant information, namely the CSI dataset, to the NW through lossless transmission.

[0452] Step 4: The NW receives the CSI dataset and, based on the training parameter information, uses the encoded CSI dataset passed through the channel and the received CSI dataset to train the second AI unit on the NW side, namely the symbol-level joint source-channel decoder, until the training termination condition is met or the training iteration number is completed.

[0453] Step 5: Establish the matching relationship between the trained UE-side symbol-level joint source-channel encoder and the NW-side symbol-level joint source-channel decoder.

[0454] In Scenario 5, the first device is the UE (User Equipment), and the second device is the NW (Network Controller). Symbol-level SSCC (Separate Training) is used. Before separate training, the UE has already trained its joint source channel encoder / decoder using the CSI dataset and preset channel information, eliminating the need for further encoder training. Furthermore, channel effects are added at the UE side of the transmitter, and the decoder is trained by the NW side. The data acquisition method for joint source channel coding in this scenario can include the following process:

[0455] Step 1: The UE uses the first AI unit, i.e., the symbol-level joint source channel encoder, to encode all samples in the first information, i.e., the CSI dataset, to obtain the second information, i.e., the encoded CSI dataset.

[0456] Step 2: The UE transmits the encoded CSI dataset through the corresponding channel according to the first channel information to obtain the encoded CSI dataset transmitted through the channel, and then sends it to the NW through lossless transmission.

[0457] Step 3: The UE sends the first relevant information, namely the CSI dataset, to the NW through lossless transmission.

[0458] Step 4: The NW receives the CSI dataset and, based on the training parameter information, uses the encoded CSI dataset passed through the channel and the received CSI dataset to train the second AI unit on the NW side, namely the symbol-level joint source-channel decoder, until the training termination condition is met or the training iteration number is completed.

[0459] Step 5: Establish the matching relationship between the trained UE-side symbol-level joint source-channel encoder and the NW-side symbol-level joint source-channel decoder.

[0460] In Scenario 6, the first device is the NW (Network Controller) and the second device is the UE (User Equipment). Symbol-level SSCC (Separate Training) is used. Before separate training, the NW has already trained its joint source channel encoder / decoder using the CSI dataset and preset channel information, eliminating the need for decoder training. The encoder is trained on the UE side, without adding channel interference. The data acquisition method for joint source channel coding in this scenario can include the following process:

[0461] Step 1: NW uses a symbol-level joint source-channel encoder to encode all samples in the first information, i.e., the CSI dataset, to obtain the second information, i.e., the encoded CSI dataset.

[0462] Step 2: The NW sends the encoded CSI dataset and the first relevant information, namely the CSI dataset, to the UE, either separately or together, via lossless transmission.

[0463] Step 3: The UE receives the encoded CSI dataset and the first relevant information. Based on the training parameter information, it uses the second CSI dataset and the encoded CSI dataset to train the second AI unit on the UE side, namely the symbol-level joint source-channel encoder, until the training termination condition is met or the training iteration number is completed.

[0464] Step 4: Establish the matching relationship between the trained UE-side symbol-level joint source-channel encoder and the NW-side symbol-level joint source-channel decoder.

[0465] The application of the above method will be explained in detail below using terminal and network-side devices as examples.

[0466] The first example: The first device is a terminal, and the second device is a network-side device. Based on the data acquisition method of joint source-channel coding described above, the process of training and application using joint training can specifically include the following steps:

[0467] Step 1: The terminal inputs the first information into the first AI unit to obtain the second information; wherein, the first information includes CSI, the first AI unit is used to encode the first information, and the second information includes the encoded CSI.

[0468] Step 2: The terminal transmits the second information to the network-side device using a preset method, wherein the preset method is used to add channel influence.

[0469] Step 3: The terminal receives the gradient information of the second information sent by the network-side device. The gradient information of the second information is obtained by the network-side device training the second AI unit based on the second information. The second AI unit is used to decode the second information.

[0470] Step 4: The terminal trains the first AI unit based on the gradient information of the second information.

[0471] Step 5: The terminal establishes a matching relationship between the first AI unit trained and the second AI unit trained by the network-side device.

[0472] Accordingly, the network-side device establishes a matching relationship between the trained second AI unit and the trained first AI unit on the terminal.

[0473] The above process is the training process. After training is completed, the training results can be used for actual CSI feedback.

[0474] Step 6: In the CSI feedback scenario, the terminal uses the trained first AI unit to encode the CSI and sends it to the network-side device. Correspondingly, the network-side device can use the trained second AI unit to decode the received encoded CSI to obtain the recovered CSI.

[0475] The aforementioned CSI feedback scenarios refer to scenarios where the terminal feeds back CSI to the network-side equipment. These scenarios can be varied, including but not limited to at least one of the following: scenarios where CSI feedback is required at the cell edge, scenarios where the bandwidth allocated for CSI feedback is relatively small, scenarios where the feedback channel is a rapidly changing channel, and scenarios where accurate channel estimation is not possible.

[0476] The above process considers the influence of the channel during AI-based coding, enabling data acquisition for AI-based joint source-channel coding. Joint training with CSI feedback enhancement based on the acquired data improves the practicality of the trained model, making it more widely applicable. Applying CSI feedback based on the trained model provides a gain to CSI, improving its accuracy and transmission efficiency.

[0477] The second example: The first device is a network-side device, and the second device is a terminal. Based on the data acquisition method of joint source-channel coding described above, the process of training and application using the separate training method can specifically include the following steps:

[0478] Step 1: The network-side device initially trains the decoder and inputs the first information into the first AI unit to obtain the second information; wherein, the first information includes CSI, and the first AI unit is used to encode the first information; the second information includes the encoded CSI.

[0479] Step 2: The network-side device transmits the second information to the terminal using a preset method, wherein the preset method is used to add channel influence.

[0480] Step 3: The terminal receives the second information sent by the network side and trains the second AI unit based on the training parameter information, the first relevant information and the second information.

[0481] The training parameter information refers to the parameters related to training the second AI unit. The first related information is information associated with the first information and the encoding / decoding, which can be sent to the terminal by the network-side device. The second AI unit is used to encode the CSI.

[0482] In one implementation, the first relevant information can be a CSI dataset, including multiple CSI samples. The network-side device can send the CSI dataset to the terminal for training on the terminal side. During terminal training, the second AI unit, the encoder, is trained using the first relevant information, i.e., the CSI dataset, as input and the second information, i.e., the encoded CSI, as labels.

[0483] Step 4: The network-side device establishes a matching relationship between the decoder and the second AI unit trained by the terminal.

[0484] Accordingly, the terminal establishes a matching relationship between the trained second AI unit and the decoder of the network-side device.

[0485] The above process is the training process. After training is completed, the training results can be used for actual CSI feedback.

[0486] Step 5: In the CSI feedback scenario, the terminal uses the trained second AI unit to encode the CSI and sends it to the network-side device. Correspondingly, the network-side device can use a decoder to decode the received encoded CSI to obtain the recovered CSI.

[0487] The aforementioned CSI feedback scenarios refer to scenarios where the terminal feeds back CSI to the network-side equipment. These scenarios can be varied, including but not limited to at least one of the following: scenarios where CSI feedback is required at the cell edge, scenarios where the bandwidth allocated for CSI feedback is relatively small, scenarios where the feedback channel is a rapidly changing channel, and scenarios where accurate channel estimation is not possible.

[0488] The above process considers the influence of the channel during AI-based coding, enabling data acquisition for AI-based joint source-channel coding. Bilateral separate training with CSI feedback enhancement based on the data acquisition results improves the practicality of the trained model, making its application more widespread. Applying CSI feedback based on the trained model can bring gains to CSI, improve the accuracy of CSI feedback, and increase transmission efficiency.

[0489] Figure 6 This illustration shows a schematic diagram of a data acquisition device with joint source-channel coding provided in an embodiment of this application, such as... Figure 6 As shown, the device 600 is applied to the first device and may include a processing module 601 and a transmission module 602.

[0490] The processing module 601 is used to input the first information into the first AI unit to obtain the second information.

[0491] The transmission module 602 is used to transmit the second information to the second device using a preset method.

[0492] The first information includes Channel State Information (CSI), the first AI unit is used to encode the first information, and the preset method is used to add channel effects.

[0493] In this embodiment, the preset method includes: lossless transmission via a channel, lossless transmission, and lossy transmission. The transmission module 602 uses the preset method to transmit the second information to the second device, which may include one of the following:

[0494] The second information is transmitted to the second device through the channel corresponding to the preset channel information using a lossless transmission method;

[0495] The second information is sent to the second device using lossless transmission.

[0496] The second information is sent to the second device using a lossy transmission method.

[0497] In this embodiment of the application, the above-mentioned device is further configured to: receive gradient information of second information sent by the second device, and train the first AI unit based on the gradient information of the second information.

[0498] The gradient information of the second information is obtained by the second device training the second AI unit based on the second information, and the second AI unit is used to decode the second information.

[0499] In one implementation, training the first AI unit based on the gradient information of the second information may include: training the first AI unit based on the gradient information of the training parameter information and the second information.

[0500] The training parameter information includes parameter information related to the training of the first AI unit.

[0501] In this embodiment of the application, CSI may include at least one of the following parameters:

[0502] Channel Quality Indicator (CQI);

[0503] Precoding matrix indicates PMI;

[0504] Channel State Information Reference Signal Resource Indicator (CRI);

[0505] Synchronization signal physical broadcast channel block resource indicator (SSBRI);

[0506] Layer indicator LI;

[0507] Rank indicator RI;

[0508] Layer 1 reference signal received power L1-RSRP;

[0509] Layer 1 signal-to-interference-plus-noise ratio (L1-SINR);

[0510] Capacity index;

[0511] Time Differential Carrier Phase (TDCP)

[0512] Raw channel information;

[0513] Transform domain channel information;

[0514] The processed raw channel information;

[0515] The information source corresponding to PMI.

[0516] In the embodiments of this application, any of the parameters included in CSI can be replaced by at least one of the mean, variance, bias and product factor of the parameter normalized, and there is no specific limitation.

[0517] In this embodiment of the application, the second information may include at least one of a bit sequence and a symbol sequence. The bit sequence corresponds to bit-level joint source-channel coding, and the symbol sequence corresponds to symbol-level joint source-channel coding.

[0518] In one implementation, the second information may include a bit sequence, and the transmission module 602 may further be configured to: modulate the bit sequence in the second information using at least one of the following modulation methods before transmitting the second information to the second device using a lossy transmission method:

[0519] Binary Phase Shift Keying (BPSK);

[0520] pi / 2-BPSK;

[0521] Quadrature Phase Shift Keying (QPSK)

[0522] 16 Quadrature Amplitude Modulation (QAM);

[0523] 64QAM;

[0524] 256QAM;

[0525] 1024QAM;

[0526] 4096QAM;

[0527] AI-based modulation.

[0528] In this embodiment of the application, the above-mentioned device is further used to: send the first relevant information to the second device using lossy transmission or lossless transmission.

[0529] Among them, the first relevant information is information that is related to the first information and the encoding / decoding.

[0530] In this embodiment of the application, the first relevant information may include at least one of the following:

[0531] Samples from the CSI dataset;

[0532] Sample IDs in the CSI dataset;

[0533] Encoded dimensional information;

[0534] Channel impact information;

[0535] The rules for obtaining the first information.

[0536] In this embodiment of the application, the rule for obtaining the first information may include an iteration count index, which is used to indicate how many samples are obtained from the current position in the samples arranged in a preset order. The current position is the initial position or the position where the last sample acquisition ended.

[0537] In this embodiment of the application, the first information can be obtained in at least one of the following ways:

[0538] Obtain the first information predefined by the protocol;

[0539] Collect first information;

[0540] Receive the first information sent by the second device;

[0541] Receive the first information sent by the server of the first device;

[0542] Receive the first message sent by the third-party server.

[0543] In this embodiment of the application, the above-mentioned device is further used to: send preset channel information to a second device using lossy or lossless transmission methods.

[0544] Among them, the preset channel information is used to add channel influence to the second information using the corresponding channel.

[0545] In this embodiment of the application, the preset channel information may include at least one of the following:

[0546] Signal-to-noise ratio (SNR);

[0547] SNR range;

[0548] Bit error rate;

[0549] Bit error rate range;

[0550] CQI;

[0551] CQI range;

[0552] Modulation and coding strategy (MCS);

[0553] MCS range;

[0554] Raw channel information of the transmission channel;

[0555] Channel model.

[0556] In this embodiment of the application, the preset channel information can be preset by the protocol, or indicated by the first device or the second device using at least one of the following methods:

[0557] Radio Resource Control (RRC) indication;

[0558] Media Access Control (MAC) control element CE indication;

[0559] Control information indication.

[0560] In this embodiment of the application, lossy transmission methods may include at least one of the following:

[0561] Layer mapping;

[0562] Intertwined;

[0563] filling;

[0564] Drilling holes;

[0565] Rate matching;

[0566] Scrambling;

[0567] Precoding.

[0568] In this embodiment of the application, the lossless transmission method may include at least one of the following:

[0569] Source coding;

[0570] Channel coding;

[0571] modulation;

[0572] Layer mapping;

[0573] Intertwined;

[0574] filling;

[0575] Drilling holes;

[0576] Rate matching;

[0577] Scrambling;

[0578] Precoding.

[0579] In this embodiment of the application, the training parameter information may include at least one of the following:

[0580] Batch size;

[0581] Optimizer type;

[0582] Learning rate;

[0583] Learning rate decay;

[0584] Loss function;

[0585] Regularization methods and parameters;

[0586] Initialization methods and parameters;

[0587] Discard the method and parameters;

[0588] Training cycle;

[0589] Number of iterations;

[0590] Training termination conditions.

[0591] In this embodiment of the application, the loss function may include:

[0592] The output of the target model is a combination function of at least one of the following: label error, mean squared error, normalized mean squared error, correlation, entropy, mutual information, and constant.

[0593] The target model is either the first AI unit or the second AI unit.

[0594] In one implementation, the loss function is obtained by weighted combination of loss information from multiple parts of the target model's output. These multiple parts are obtained by partitioning according to at least one of spatial domain resources, code domain resources, frequency domain resources, and time domain resources. Frequency domain resources include resource blocks (RBs), subbands, or precoding resource block groups (PRGs). Time domain resources include subcarriers, symbols, time slots, or half-time slots. The loss information includes the loss value and / or a function relating the loss.

[0595] In this embodiment of the application, the training parameter information may be preset by the protocol, or indicated by the first device or the second device using at least one of the following methods:

[0596] RRC indication;

[0597] MAC CE indication;

[0598] Control information indication.

[0599] In this embodiment of the application, the second information may be contained in one of the following:

[0600] DCI;

[0601] MAC CE;

[0602] RRC;

[0603] Physical Uplink Shared Channel (PUSCH);

[0604] Physical Downlink Shared Channel (PDSCH);

[0605] Physical Downlink Control Channel (PDCCH);

[0606] Physical uplink control channel (PUCCH).

[0607] In this embodiment, the second information can be configured by RRC or MAC CE to be sent periodically, semi-persistently, or aperiodically. Semi-persistent transmission is activated by MAC CE or DCI and is sent periodically after activation until deactivated or after a specified period. Aperiodic transmission is activated by MAC CE or DCI and is sent a specified number of times.

[0608] In this embodiment of the application, the gradient information of the second information can be carried in one of the following ways:

[0609] DCI;

[0610] MAC CE;

[0611] RRC;

[0612] PUSCH;

[0613] PDSCH;

[0614] PDCCH;

[0615] PUCCH.

[0616] In this embodiment, the gradient information of the second information can be configured by RRC or MAC CE to be sent periodically, semi-persistently, or aperiodically. Semi-persistent transmission is activated by MAC CE or DCI and is sent periodically after activation until deactivated or after a specified period. Aperiodic transmission is activated by MAC CE or DCI and is sent a specified number of times.

[0617] In this embodiment of the application, the gradient information of the second information can be the gradient information of the second information that is greater than or equal to a preset first threshold.

[0618] In this embodiment of the application, the above-mentioned device is further used to: establish a matching relationship between the first AI unit and the second AI unit after training.

[0619] In this embodiment of the application, the identifier of the matching relationship may include at least one of the following:

[0620] Model ID;

[0621] Function ID;

[0622] Dataset ID;

[0623] Matching ID;

[0624] Associated ID.

[0625] In this embodiment of the application, the training of the first AI unit or the training of the second AI unit can be triggered by one of the following:

[0626] Network instructions;

[0627] Terminal reporting;

[0628] Predefined events.

[0629] In this embodiment of the application, obtaining the second information can be triggered by one of the following:

[0630] Network instructions;

[0631] Terminal reporting;

[0632] Predefined events.

[0633] In this embodiment of the application, the acquisition of the first relevant information can be triggered by one of the following:

[0634] Network instructions;

[0635] Terminal reporting;

[0636] Predefined events.

[0637] In this embodiment of the application, the predefined event may include at least one of the following:

[0638] Currently in Wi-Fi mode;

[0639] The battery level has exceeded the second threshold.

[0640] Currently charging;

[0641] Supports preset traffic patterns;

[0642] In preset traffic mode;

[0643] No business need;

[0644] The performance index is below the third threshold.

[0645] In this embodiment of the application, the second information can be used by the second device to train the second AI unit.

[0646] In this embodiment of the application, the first AI unit and the second AI unit can align third information before training. The third information includes at least one of the following: structural features of the model, load quantization method of the model, estimation accuracy of the model, and output accuracy of the model.

[0647] The apparatus provided in this application embodiment can execute the method in any of the method embodiments with the first device as the execution subject. For details, please refer to the description in the method embodiments, which will not be repeated here.

[0648] The apparatus provided in this application embodiment inputs first information into a first AI unit to obtain second information through a first device, and transmits the second information to a second device using a preset method. The first information is CSI, the first AI unit is used to encode the first information, and the preset method is used to add channel effects, so that the second device can obtain the second information with added channel effects. It can consider the channel effects in the AI-based encoding process, realize the data acquisition of AI-based joint source-channel coding, help CSI feedback enhancement training, improve the practicality of the training model, and make the model more widely used.

[0649] Figure 7 This illustration shows a schematic diagram of a data acquisition device with joint source-channel coding provided in an embodiment of this application, such as... Figure 7As shown, the device 700 is applied to a second device and may include a receiving module 701.

[0650] The receiving module 701 is used to receive second information sent by the first device using a preset method.

[0651] The preset method is used to add channel influence. The second information is obtained by the first device inputting the first information into the first AI unit. The first information includes CSI. The first AI unit is used to encode the first information. The second information includes the encoded CSI.

[0652] In this embodiment of the application, the above-mentioned device is also used to: train a second AI unit based on the second information.

[0653] The aforementioned device, which trains a second AI unit based on second information, may include:

[0654] The second information is input into the second AI unit for decoding to obtain the recovered first information;

[0655] Based on the training parameter information, the first relevant information, and the recovered first information, the second AI unit is trained to obtain the gradient information of the second information.

[0656] In this embodiment of the application, the receiving module 701 receives second information sent by the first device using a preset method, which may include one of the following:

[0657] The first device receives the second information transmitted through the corresponding channel according to the preset channel information using a lossless transmission method.

[0658] Receive the second information sent by the first device using lossless transmission;

[0659] Receive the second information sent by the first device using lossy transmission.

[0660] In this embodiment of the application, the above-mentioned device can also be used to: after receiving the second information sent by the first device using lossless transmission, train the second AI unit by passing the second information through the corresponding channel according to the preset channel information.

[0661] In this embodiment of the application, the above-described device can also be used to: send the gradient information of the second information to the first device.

[0662] The gradient information of the second information is used by the first device to train the first AI unit in combination with the training parameter information.

[0663] In this embodiment of the application, the above-described device can also be used to: receive first relevant information sent by the first device using lossy transmission or lossless transmission.

[0664] Among them, the first relevant information is information that is related to the first information and the encoding / decoding.

[0665] In this embodiment of the application, the above-mentioned device can also be used to: receive preset channel information transmitted by the first device using lossy transmission or lossless transmission.

[0666] Among them, the preset channel information is used to add channel effects to the second information using the corresponding channel.

[0667] In this embodiment of the application, the above-mentioned preset method is a lossy transmission method. After the receiving module 701 receives the second information sent by the first device using the preset method, it further includes performing at least one of the following operations on the second information:

[0668] Demapping;

[0669] Untangling;

[0670] Remove fill;

[0671] Disturbing;

[0672] Receive and merge.

[0673] In this embodiment of the application, the preset method includes a lossless transmission method. After receiving the second information sent by the first device using the preset method, the receiving module 701 further includes performing at least one of the following operations on the second information:

[0674] Source decoding;

[0675] Channel decoding;

[0676] demodulation;

[0677] Demapping;

[0678] Untangling;

[0679] Remove fill;

[0680] Disturbing;

[0681] Receive and merge.

[0682] The apparatus provided in this application embodiment can execute the method in any of the method embodiments with the second device as the execution subject. For details, please refer to the description in the method embodiments, which will not be repeated here.

[0683] The apparatus provided in this application embodiment receives second information sent by a first device using a preset method through a second device. The preset method is used to add channel influence. The second information is obtained by the first device by inputting the first information into a first AI unit. The first information is CSI. The first AI unit is used to encode the first information so that the second device can obtain the second information with added channel influence. It can consider the influence of the channel in the AI-based encoding process, realize the data acquisition of AI-based joint source-channel coding, help the training of CSI feedback enhancement, improve the practicality of the training model, and make the model more widely used.

[0684] This application provides a data acquisition device for joint source-channel coding. As an example, the device can be a communication device or a component within a communication device, such as a chip. The communication device can be a terminal, a network-side device, or a server, etc. Exemplarily, the terminal can include, but is not limited to, the types of terminals listed above, and the network-side device can include, but is not limited to, the types of network-side devices listed above. This application does not impose specific limitations on these types.

[0685] The data acquisition device for joint source-channel coding includes a receiving module, a transmitting module, and a processing module. These modules can be implemented in software or hardware. When implemented in hardware, the processing module can be implemented by a processor. For example, the processor can include general-purpose processors, special-purpose processors, such as a Central Processing Unit (CPU), microprocessor, Digital Signal Processor (DSP), Artificial Intelligence (AI) processor, Graphics Processing Unit (GPU), Application Specific Integrated Circuit (ASIC), Network Processor (NP), Field Programmable Gate Array (FPGA), or other programmable logic devices, gate circuits, transistors, discrete hardware components, etc. The receiving and transmitting modules can be implemented by a communication interface, which can include one or more of the following: transceiver, pins, circuits, bus, radio frequency unit, etc.

[0686] The data acquisition device for joint source-channel coding provided in this application embodiment can achieve... Figures 2 to 5 The various processes implemented in the method embodiments achieve the same technical effect, and will not be described again here to avoid repetition.

[0687] like Figure 8 As shown in the illustration, this application also provides a communication device 800, including a processor 801 and a memory 802. The memory 802 stores programs or instructions that can run on the processor 801. For example, when the communication device 800 is a terminal, the program or instructions executed by the processor 801 implement the various steps of the above-described method embodiments and achieve the same technical effect. When the communication device 800 is a network-side device, the program or instructions executed by the processor 801 implement the various steps of the above-described method embodiments and achieve the same technical effect. To avoid repetition, further details are omitted here.

[0688] This application also provides a terminal, including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the steps in the above method embodiments. This terminal embodiment corresponds to the above-described terminal-side method embodiments; all implementation processes and methods of the above method embodiments can be applied to this terminal embodiment and achieve the same technical effect. The terminal can be... Figure 6 Or the device shown in Figure 7. Specifically, Figure 9 A schematic diagram of the hardware structure of a terminal to implement an embodiment of this application.

[0689] The terminal 900 includes, but is not limited to, at least some of the following components: radio frequency unit 901, network module 902, audio output unit 903, input unit 904, sensor 905, display unit 906, user input unit 907, interface unit 908, memory 909, and processor 910.

[0690] Those skilled in the art will understand that the terminal 900 may also include a power supply (such as a battery) for supplying power to various components. The power supply may be logically connected to the processor 910 through a power management system, thereby enabling functions such as managing charging, discharging, and power consumption through the power management system. Figure 9 The terminal structure shown does not constitute a limitation on the terminal. The terminal may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here.

[0691] It should be understood that, in this embodiment, the input unit 904 may include a graphics processor 9041 and a microphone 9042. The graphics processor 9041 processes image data of still images or videos obtained by an image capture device (such as a camera) in video capture mode or image capture mode. The display unit 906 may include a display panel 9061, which may be configured in the form of a liquid crystal display, an organic light-emitting diode, or the like. The user input unit 907 includes at least one of a touch panel 9071 and other input devices 9072. The touch panel 9071 is also called a touch screen. The touch panel 9071 may include a touch detection device and a touch controller. Other input devices 9072 may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, and joysticks, which will not be described in detail here.

[0692] In this embodiment, after receiving downlink data from the network-side device, the radio frequency unit 901 can transmit it to the processor 910 for processing; in addition, the radio frequency unit 901 can send uplink data to the network-side device. Typically, the radio frequency unit 901 includes, but is not limited to, antennas, amplifiers, transceivers, couplers, low-noise amplifiers, duplexers, etc.

[0693] The memory 909 can be used to store software programs or instructions, as well as various data. The memory 909 may primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area may store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, the memory 909 may include volatile memory or non-volatile memory. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DRRAM). The memory 909 in the embodiments of this application includes, but is not limited to, these and any other suitable types of memory.

[0694] Processor 910 may include one or more processing units; optionally, processor 910 integrates an application processor and a modem processor, wherein the application processor mainly handles operations involving the operating system, user interface, and applications, and the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into processor 910.

[0695] The processor 910 is used to input first information into a first artificial intelligence (AI) unit to obtain second information. The radio frequency (RF) unit 901 is used to transmit the second information to a second device using a preset method. The preset method is used to add channel interference. The second information is obtained by the first device inputting the first information into the first AI unit. The first information includes CSI (Content Sequence Index). The first AI unit is used to encode the first information.

[0696] Alternatively, the radio frequency unit 901 is used to receive second information transmitted by the first device using a preset method. The preset method is used to add channel effects, and the second information is obtained by the first device inputting first information into the first AI unit. The first information includes CSI, and the first AI unit is used to encode the first information.

[0697] The terminal provided in this application embodiment obtains second information by inputting first information into a first AI unit, and transmits the second information to a second device using a preset method. The first information is CSI, the first AI unit is used to encode the first information, and the preset method is used to add channel influence, so that the second device can obtain the second information with added channel influence. It can consider the influence of the channel in the AI-based encoding process, realize the data acquisition of AI-based joint source-channel coding, help CSI feedback enhancement training, improve the practicality of the training model, and make the model more widely used.

[0698] It is understood that the implementation process of each implementation method mentioned in this embodiment can refer to the relevant description of the method embodiment and achieve the same or corresponding technical effect. To avoid repetition, it will not be described again here.

[0699] This application also provides a network-side device, including a processor and a communication interface. The communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the steps of the above-described method embodiments. This network-side device embodiment corresponds to the above-described network-side device method embodiments. All implementation processes and methods of the above-described method embodiments can be applied to this network-side device embodiment and achieve the same technical effects.

[0700] Specifically, embodiments of this application also provide a network-side device, which can be... Figure 6 Or the device shown in Figure 7. (e.g.) Figure 10 As shown, the network-side device 1000 includes: an antenna 101, a radio frequency (RF) device 102, a baseband device 103, a processor 104, and a memory 105. The antenna 101 is connected to the RF device 102. In the uplink direction, the RF device 102 receives information through the antenna 101 and transmits the received information to the baseband device 103 for processing. In the downlink direction, the baseband device 103 processes the information to be transmitted and sends it to the RF device 102. The RF device 102 processes the received information and transmits it through the antenna 101.

[0701] The method executed by the network-side device in the above embodiments can be implemented in the baseband device 103, which includes a baseband processor.

[0702] The baseband device 103 may include, for example, at least one baseband board on which multiple chips are disposed, such as... Figure 10 As shown, one of the chips is, for example, a baseband processor, which is connected to the memory 105 via a bus interface to call the program in the memory 105 and execute the network device operation shown in the above method embodiment.

[0703] The network-side device may also include a network interface 106, such as a Common Public Radio Interface (CPRI).

[0704] Specifically, the network-side device 1000 in this application embodiment further includes: instructions or programs stored in memory 105 and executable on processor 104, wherein processor 104 calls the instructions or programs in memory 105 to execute. Figure 6 Alternatively, the methods executed by each module shown in Figure 7 can achieve the same technical effect. To avoid repetition, they will not be described in detail here.

[0705] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0706] The processor mentioned above is the processor in the terminal described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk. In some examples, the readable storage medium may be a non-transient readable storage medium.

[0707] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above method embodiments and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0708] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0709] This application also provides a computer program / program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the above method embodiments and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0710] This application also provides a data acquisition system using joint source-channel coding, comprising: a terminal and a network-side device. The terminal can be used to execute the steps of the first device-side joint source-channel coding data acquisition method described above, and the network-side device can be used to execute the steps of the second device-side joint source-channel coding data acquisition method described above. Alternatively, the network-side device can be used to execute the steps of the first device-side joint source-channel coding data acquisition method described above, and the terminal can be used to execute the steps of the second device-side joint source-channel coding data acquisition method described above.

[0711] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0712] From the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of computer software products plus necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, magnetic disk, optical disk, etc.) and includes several instructions to cause the terminal or network-side device to execute the methods described in the various embodiments of this application.

[0713] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other implementations under the guidance of this application without departing from the spirit and scope of the claims. All of these implementations are within the protection scope of this application.

Claims

1. A data acquisition method for joint source-channel coding, characterized by, Comprise: The first device inputs the first information into the first artificial intelligence (AI) unit to obtain second information; The second information is transmitted to the second device using a preset method; Wherein, the first information includes channel state information (CSI), the first AI unit is used for encoding the first information, and the preset method is used for adding channel influence.

2. The method of claim 1, wherein, The preset method includes: a lossless transmission method through a channel, a lossless transmission method, and a lossy transmission method.

3. The method of claim 1, wherein, Also include: Receiving gradient information of the second information sent by the second device, wherein the gradient information of the second information is obtained by the second device according to training of the second AI unit on the second information, and the second AI unit is used for decoding the second information; Training the first AI unit according to the gradient information of the second information.

4. The method according to any one of claims 1-3, characterized in that, The CSI includes at least one of the following parameters: Channel quality indicator (CQI); Precoding matrix indicator (PMI); Channel state information reference signal resource indicator (CRI); Synchronization signal physical broadcast channel block resource indicator (SSBRI); Layer indicator (LI); Rank indicator (RI); Layer 1 reference signal received power (L1-RSRP); Layer 1 signal to interference plus noise ratio (L1-SINR); Capacity index; Time difference carrier phase (TDCP); Naked channel information; Transform domain channel information; Processed original channel information; The source corresponding to the PMI.

5. The method of claim 1, wherein, The second information includes at least one of a bit sequence and a symbol sequence; wherein the bit sequence corresponds to a bit-level joint source channel encoding, and the symbol sequence corresponds to a symbol-level joint source channel encoding.

6. The method of claim 2, wherein, The second information includes a bit sequence, and before the second information is sent to the second device using the lossy transmission method, the following at least one modulation method is used to modulate the bit sequence in the second information: Binary phase shift keying (BPSK); pi / 2-BPSK; Quadrature phase shift keying (QPSK); 16 quadrature amplitude modulation (QAM); 64QAM; 256QAM; 1024QAM; 4096QAM; AI-based modulation.

7. The method according to any one of claims 1 to 6, characterized in that, Also include: Using a lossy transmission or lossless transmission method, send the first related information to the second device; wherein the first related information is information related to the first information and encoding and decoding; The first related information includes at least one of the following: Sample in the CSI dataset; Sample ID in the CSI dataset; Encoded dimension information; Channel influence information; The acquisition rule of the first information.

8. The method of claim 7, wherein, The acquisition rule includes an iteration number index, which is used to indicate that a preset number of samples are obtained from the current position in the samples arranged in a preset order, and the current position is an initial position or a position where the last sample acquisition ends.

9. The method according to any one of claims 1-8, characterized in that, The first device is a terminal or a network side device, and the first information is obtained in the following at least one way: Obtain the first information predefined by the protocol; Collect the first information; Receive the first information sent by the second device; Receive the first information sent by the server of the first device; wherein the server of the first device is a server belonging to the same network operator as the first device; Receiving the first information sent by the third-party server.

10. The method of claim 1, wherein, Further comprising: Sending preset channel information to the second device in a lossy transmission or a lossless transmission manner, wherein the preset channel information is used to add channel influence on the second information using a corresponding channel; The preset channel information comprises at least one of: Signal-to-noise ratio (SNR); SNR range; Bit error rate; Bit error rate range; CQI; CQI range; Modulation and coding strategy (MCS); MCS range; Naked channel information of a transmission channel; Channel model.

11. The method of claim 2 or 10, wherein, The preset channel information is preset by a protocol or indicated by the first device or the second device using at least one of the following manners: Radio resource control (RRC) indication; Media access control control element (MAC CE) indication; Control information indication.

12. The method of claim 2, 7, or 10, wherein, The lossy transmission manner comprises at least one of: Layer mapping; Interleaving; Padding; Puncturing; Rate matching; Scrambling; Precoding.

13. The method of claim 2, 7, or 10, wherein, The lossless transmission manner comprises at least one of: Source coding; Channel coding; Modulation; Layer mapping; Interleaving; Padding; Puncturing; Rate matching; Scrambling; Precoding.

14. The method of claim 3, wherein, The training parameter information of the first AI unit is preset by a protocol or indicated by the first device or the second device using at least one of the following manners: RRC indication; MAC CE indication; Control information indication.

15. The method of claim 1, wherein, The second information is carried in one of the following: DCI; MAC CE; RRC; Physical uplink shared channel (PUSCH); Physical downlink shared channel (PDSCH); Physical downlink control channel (PDCCH); Physical uplink control channel (PUCCH).

16. The method of claim 1, wherein, The second information is configured by RRC or MAC CE to be periodically transmitted, semi-persistently transmitted, or aperiodically transmitted; The semi-persistent transmission is activated by MAC CE or DCI, and is periodically transmitted after activation until deactivation or after a specified period; The aperiodic transmission is activated by MAC CE or DCI, and is transmitted a specified number of times.

17. The method of claim 3, wherein, The gradient information of the second information is carried in one of the following: DCI; MAC CE; RRC; PUSCH; PDSCH; PDCCH; PUCCH.

18. The method of claim 3, wherein, The gradient information of the second information is configured by RRC or MAC CE to be periodically transmitted, semi-persistently transmitted, or aperiodically transmitted; The semi-persistent transmission is activated by MAC CE or DCI, and is periodically transmitted after activation until deactivation or after a specified period; The aperiodic transmission is activated by MAC CE or DCI, and is transmitted a specified number of times.

19. The method of claim 3, wherein, The gradient information of the second information is greater than or equal to a preset first threshold.

20. The method of claim 3, wherein, Further comprising: Establishing a matching relationship between the trained first AI unit and the second AI unit.

21. The method of claim 20, wherein, The identifier of the matching relationship comprises at least one of: Model ID; Function ID; Dataset ID; Pairing ID; Association ID.

22. The method of any one of claims 3, 17-21, wherein, The training of the first AI unit or the training of the second AI unit is triggered by one of the following: Network indication; Terminal reporting; Predefined event.

23. The method of claim 1, wherein, The obtaining of the second information is triggered by one of the following: Network indication; Terminal reporting; Predefined event.

24. The method of claim 7, wherein, The obtaining of the first related information is triggered by one of the following: Network indication; Terminal reporting; Predefined event.

25. The method of claim 22, 23, or 24, wherein, The predefined event includes at least one of the following: In a wifi state; Battery power exceeding a second threshold; In a charging state; Supporting a preset traffic mode; In a preset traffic mode; No service demand; Performance index below a third threshold.

26. The method of any one of claims 1-25, wherein, The second information is used for training the second AI unit by the second device.

27. The method of claim 3, wherein, The first AI unit and the second AI unit are aligned with third information before training, and the third information includes at least one of the following: structural features of a model, load quantization method of a model, estimation accuracy of a model, and output accuracy of a model.

28. A data acquisition method for joint source-channel coding, characterized by, It includes: The second device receives second information sent by the first device in a preset manner; Wherein, the preset manner is used to add channel influence, the second information is obtained by inputting the first information into the first AI unit by the first device, and the first information includes CSI, and the first AI unit is used for encoding the first information.

29. The method of claim 28, wherein, It also includes: Inputting the second information into the second AI unit for decoding to obtain recovered first information; According to the training parameter information, the first related information and the recovered first information, training the second AI unit to obtain the gradient information of the second information; The gradient information of the second information is sent to the first device, and the gradient information of the second information is used for the first device to train the first AI unit in combination with the training parameter information.

30. The method of claim 28, wherein, The preset manner includes: a lossless transmission manner through a channel, a lossless transmission manner, and a lossy transmission manner, and the second device receives the second information sent by the first device in the preset manner, including one of the following: The second device receives the second information sent by the first device in a lossless transmission manner through a corresponding channel according to preset channel information; The second device receives the second information sent by the first device in a lossless transmission manner, and inputs the second information into the corresponding channel according to the preset channel information; The second device receives the second information sent by the first device in a lossy transmission manner.

31. The method of any one of claims 28-30, wherein, It also includes: Receiving the first related information sent by the first device in a lossy transmission or lossless transmission manner; wherein, the first related information is information related to the first information and encoding and decoding; The first related information includes at least one of the following: Sample in CSI data set; Sample ID in CSI data set; Encoded dimension information; Channel influence information; First information acquisition rule.

32. The method of any one of claims 28-30, wherein, It also includes: Receiving preset channel information sent by the first device in a lossy transmission or lossless transmission manner; wherein, the preset channel information is used to add channel influence to the second information using a corresponding channel; The preset channel information includes at least one of the following: Signal-to-noise ratio (SNR); SNR range; Bit error rate; Bit error rate range; CQI; CQI range; Modulation and coding strategy (MCS); MCS range; Raw channel information of transmission channel; Channel model.

33. The method of any one of claims 30-32, wherein, The preset manner is a lossy transmission manner, and after the second device receives the second information sent by the first device in the preset manner, it further includes at least one of the following operations on the second information: Demapping; Deinterleaving; Depopulation; descrambling; receive combining.

34. The method of any one of claims 30-32, wherein, The preset manner includes a lossless transmission manner, and after the second device receives second information sent by the first device using the preset manner, the second device further performs at least one of the following operations on the second information: source coding; channel coding; demodulation; de-mapping; de-interleaving; de-padding; descrambling; receive combining.

35. A data acquisition device incorporating joint source-channel coding, characterized by comprising: a processing module, configured to input first information into a first artificial intelligence (AI) unit to obtain second information; a transmission module, configured to transmit the second information to a second device using a preset manner; wherein the first information includes channel state information (CSI), the first AI unit is configured to encode the first information, and the preset manner is configured to add channel influence.

36. A data acquisition device incorporating joint source-channel coding, characterized by comprising: a receiving module, configured to receive second information sent by a first device using a preset manner; wherein the preset manner is configured to add channel influence, the second information is obtained by inputting first information into a first AI unit by the first device, the first information includes CSI, and the first AI unit is configured to encode the first information.

37. A terminal, characterized by a processor and a memory, the memory stores programs or instructions executable on the processor, and the programs or instructions are executed by the processor to implement the steps of the joint source-channel coding data acquisition method according to any one of claims 1-34.

38. A network-side device, comprising: a processor and a memory, the memory stores programs or instructions executable on the processor, and the programs or instructions are executed by the processor to implement the steps of the joint source-channel coding data acquisition method according to any one of claims 1-34.

39. A readable storage medium characterized by, programs or instructions stored on the readable storage medium, and the programs or instructions are executed by the processor to implement the steps of the joint source-channel coding data acquisition method according to any one of claims 1-34.

Images