Information feedback method and related apparatus

Abstract

The application discloses an information feedback method and related devices, the method comprises: a terminal device performs downlink channel estimation to obtain a channel matrix; the terminal device compresses the channel matrix based on a first processing sub-model to obtain a compressed channel matrix; wherein the first processing sub-model is determined by a processing model, the segmentation point corresponding to the terminal device and the part before the segmentation point in the processing model are the first processing sub-model, the part after the segmentation point in the processing model is a second processing sub-model, and the second processing sub-model is used for decompressing the compressed channel matrix; the terminal device sends channel state information to a network device, the channel state information comprises the compressed channel matrix and the identification of the segmentation point, and the identification of the segmentation point is used for indicating the segmentation point; the network device decompresses the compressed channel matrix based on the second processing sub-model to obtain the channel matrix; and the second processing sub-model is determined by the segmentation point. The application can make the terminal device flexibly feedback the channel state information.

Description

Technical Field

[0001] This application relates to the field of communication technologies, and in particular, to an information feedback method and related devices. Background Art

[0002] A multiple-input multiple-output (MIMO) system improves the capacity and spectral efficiency of a communication system by deploying multiple antennas on the base station side (or the network device side) and the terminal device side to form multiple channels between the transmitter and the receiver. In a MIMO system, the base station can communicate with different terminal devices through multiple antennas. To eliminate interference between terminal devices, the base station needs to perform precoding on a message and then send the message to a specific terminal device on the premise of knowing the channel state information (CSI) of the downlink channel. In a specific communication scenario, the CSI is determined by the terminal device and fed back to the base station through an uplink control channel or an uplink shared channel. Traditional CSI feedback is based on a codebook. In this method, the terminal device first determines a precoding matrix from the codebook according to the estimation result of the downlink channel, and then feeds back the indication information corresponding to the precoding matrix to the base station as the CSI. This method will have complex codebook design and large feedback overhead as the number of antennas increases. Therefore, a method of compressing the CSI based on artificial intelligence (AI) technology and then feeding it back to the base station has been proposed currently.

[0003] See Figure 1 , which is a CSI feedback method based on AI technology provided by an embodiment of this application. As Figure 1 shown, first, the terminal device estimates the channel based on the channel state information reference signal CSI-RS to obtain a channel estimation result (i.e., a channel matrix), and then compresses the channel matrix through an AI encoder to obtain binary bits; when the base station receives the binary bits sent by the terminal device, the base station uses an AI decoder配套 with the AI encoder to restore the binary bits to a channel matrix. In this method, the input and output data sizes (dimensions) of the AI encoder and the AI decoder are relatively fixed, which makes it less flexible for the terminal device to feed back the CSI to the network device and the feedback efficiency is low. Summary of the Invention

[0004] This application provides an information feedback method and related devices, which can enable a terminal device to flexibly feed back channel state information to a network device and improve the feedback efficiency of the channel state information.

[0005] In a first aspect, this application provides an information feedback method applied to a terminal device. The method includes: performing downlink channel estimation to obtain a channel matrix; compressing the channel matrix based on a first processing sub-model to obtain a compressed channel matrix; wherein the first processing sub-model is determined according to a processing model, the segmentation point corresponding to the terminal device and the portion before the segmentation point in the processing model constitute the first processing sub-model, and the portion after the segmentation point in the processing model constitutes a second processing sub-model, the second processing sub-model being used to decompress the compressed channel matrix; and sending channel state information to a network device, the channel state information including the compressed channel matrix and the identifier of the segmentation point, the identifier of the segmentation point being used to indicate the segmentation point.

[0006] In one possible implementation, the segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

[0007] In one possible implementation, the processing model includes at least one processing layer, the processing model includes at least one segmentation point, a segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point; the segmentation point corresponding to the terminal device in the processing model and the portion before the segmentation point constitute the first processing sub-model, including: the segmentation point corresponding to the terminal device in the processing model and the processing layer before the segmentation point constitute the first processing sub-model.

[0008] In one possible implementation, the segmentation point is determined based on segmentation point information and first information, wherein the segmentation point information includes the identifier of at least one segmentation point of the processing model; the first information includes one or more of the following: capability information of the terminal device, channel quality of the channel between the terminal device and the network device; the capability information includes one or more of the following: computing power, communication power, and power information.

[0009] In one possible implementation, the segmentation point information further includes the dimension information and computational complexity information of the at least one segmentation point; the dimension information of a segmentation point is used to indicate the dimension of the data output by the segmentation point, and the computational complexity information of a segmentation point is used to indicate the computational complexity required for the processing sub-model corresponding to the segmentation point to run; the processing sub-model corresponding to a segmentation point is the segmentation point and the part before the segmentation point in the processing model.

[0010] In one possible implementation, the method further includes: receiving the segmentation point information from the network device.

[0011] In one possible implementation, the dimensional information and computational complexity information of the at least one segmentation point are determined based on the identifier of the at least one segmentation point and the processing model.

[0012] In one possible implementation, the method further includes: receiving the processing model from the network device.

[0013] Secondly, this application provides another information feedback method applied to a network device. The method includes: receiving channel state information, the channel state information including a channel matrix compressed using a first processing sub-model and identifiers of segmentation points corresponding to terminal devices; wherein the first processing sub-model is determined according to a processing model, the identifiers of the segmentation points are used to indicate the segmentation points, the segmentation points and the portion before the segmentation points in the processing model constitute the first processing sub-model, the portion after the segmentation points in the processing model constitutes a second processing sub-model, the second processing sub-model is used to decompress the compressed channel matrix; decompressing the compressed channel matrix based on the second processing sub-model to obtain the channel matrix; the second processing sub-model is determined according to the segmentation points.

[0014] In one possible implementation, the segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

[0015] In one possible implementation, the processing model includes at least one processing layer, the processing model includes at least one segmentation point, a segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point; the portion of the processing model after the segmentation point is a second processing sub-model, including: the processing layer after the segmentation point in the processing model is the second processing sub-model.

[0016] In one possible implementation, the at least one segmentation point is determined based on the dimensional information and computational complexity information of each processing layer in the processing model; the dimensional information of a processing layer is used to indicate the dimension of the data output by the processing layer, and the computational complexity information of a processing layer is used to indicate the computational complexity required to run the processing sub-model corresponding to the processing layer; the processing sub-model corresponding to a processing layer is the processing layer and the part before the processing layer in the processing model.

[0017] In one possible implementation, the maximum number of the at least one segmentation point is determined by the number of bits occupied by the identifier of the segmentation point.

[0018] In one possible implementation, the method further includes: sending the processing model and segmentation point information to the terminal device, wherein the segmentation point information includes the identifier of the at least one segmentation point.

[0019] In one possible implementation, the segmentation point information further includes the dimension information and computational cost information corresponding to the at least one segmentation point.

[0020] Thirdly, this application provides a communication device, including units for implementing the method of the first aspect and any possible implementation thereof, or units for implementing the method of the second aspect and any possible implementation thereof.

[0021] Fourthly, this application provides a communication device, including a processor and a transceiver; the transceiver is used to receive or transmit signals; the processor is used to perform the method as described in the first aspect above and any possible implementation thereof, or to perform the method as described in the second aspect above and any possible implementation thereof.

[0022] In one possible implementation, the communication device further includes a memory for storing a computer program; the processor is specifically configured to invoke the computer program from the memory, causing the communication device to perform the method as described in the first aspect and any possible implementation thereof, or to perform the method as described in the second aspect and any possible implementation thereof.

[0023] Fifthly, this application provides a chip for performing downlink channel estimation to obtain a channel matrix; the chip is further configured to compress the channel matrix based on a first processing sub-model to obtain a compressed channel matrix; wherein the first processing sub-model is determined according to a processing model, the segmentation point corresponding to the terminal device in the processing model and the part before the segmentation point constitute the first processing sub-model, and the part after the segmentation point in the processing model constitutes a second processing sub-model, the second processing sub-model being used to decompress the compressed channel matrix; the chip is further configured to send channel state information to a network device, the channel state information including the compressed channel matrix and the identifier of the segmentation point, the identifier of the segmentation point being used to indicate the segmentation point.

[0024] Sixthly, this application provides a chip for receiving channel state information, the channel state information including a channel matrix compressed using a first processing sub-model and identifiers of segmentation points corresponding to terminal devices; wherein the first processing sub-model is determined according to a processing model, the identifiers of the segmentation points are used to indicate the segmentation points, the segmentation points and the portion before the segmentation points in the processing model constitute the first processing sub-model, the portion after the segmentation points in the processing model constitutes a second processing sub-model, and the second processing sub-model is used to decompress the compressed channel matrix; the chip is further used to decompress the compressed channel matrix based on the second processing sub-model to obtain the channel matrix; the second processing sub-model is determined according to the segmentation points.

[0025] In a seventh aspect, this application provides a module device, the module device including a communication module, a power module, a storage module, and a chip module, wherein: the power module is used to provide power to the module device; the storage module is used to store data and instructions; the communication module is used for internal communication within the module device, or for communication between the module device and external devices; the chip module is used for: performing downlink channel estimation to obtain a channel matrix; compressing the channel matrix based on a first processing sub-model to obtain a compressed channel matrix; wherein the first processing sub-model is determined according to a processing model, the segmentation point corresponding to the terminal device in the processing model and the part before the segmentation point are the first processing sub-model, the part after the segmentation point in the processing model are the second processing sub-model, the second processing sub-model is used to decompress the compressed channel matrix; and sending channel state information to a network device, the channel state information including the compressed channel matrix and the identifier of the segmentation point, the identifier of the segmentation point being used to indicate the segmentation point.

[0026] Eighthly, this application provides a module device, the module device including a communication module, a power module, a storage module, and a chip module, wherein: the power module is used to provide power to the module device; the storage module is used to store data and instructions; the communication module is used for internal communication within the module device, or for communication between the module device and external devices; the chip module is used to: receive channel state information, the channel state information including a channel matrix compressed using a first processing sub-model and an identifier of a segmentation point corresponding to a terminal device; wherein, the first processing sub-model is determined according to a processing model, the identifier of the segmentation point is used to indicate the segmentation point, the segmentation point and the portion before the segmentation point in the processing model constitute the first processing sub-model, the portion after the segmentation point in the processing model constitutes a second processing sub-model, the second processing sub-model is used to decompress the compressed channel matrix; decompress the compressed channel matrix based on the second processing sub-model to obtain the channel matrix; the second processing sub-model is determined according to the segmentation point.

[0027] Ninthly, this application provides a computer-readable storage medium storing computer-readable instructions that, when executed on a communication device, cause the communication device to perform the method as described in the first aspect and any possible implementation thereof, or to perform the method as described in the second aspect and any possible implementation thereof.

[0028] In a tenth aspect, this application provides a computer program or computer program product, including code or instructions that, when executed on a computer, cause the computer to perform the method as described in the first aspect and any possible implementation thereof, or to perform the method as described in the second aspect and any possible implementation thereof.

[0029] This application embodiment divides the processing model into a first processing model and a second processing model based on a segmentation point corresponding to the terminal device. The terminal device uses the first processing model to compress the channel matrix, and the network device uses the second processing model to recover the compressed channel matrix. During this process, the selection of the segmentation point is related to the terminal device, and the first and second processing models change with the segmentation point. The dimension (or size) of the compressed channel matrix obtained through the first processing model also changes accordingly. Therefore, the terminal device in this application can flexibly feed back channel state information to the network device, improving the efficiency of channel state information feedback. Attached Figure Description

[0030] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a flowchart illustrating a channel state information feedback method based on artificial intelligence technology provided in an embodiment of this application;

[0032] Figure 2 This is a schematic diagram of a communication system architecture for information feedback provided in an embodiment of this application;

[0033] Figure 3 This is a flowchart illustrating an information feedback method provided in an embodiment of this application;

[0034] Figure 4 This is an example diagram of information feedback provided in an embodiment of this application;

[0035] Figure 5 This is a flowchart illustrating another information feedback method provided in an embodiment of this application;

[0036] Figure 6 This is a bar chart illustrating the relevant information of each processing layer in a processing model provided in this application embodiment;

[0037] Figure 7 This is a flowchart illustrating another information feedback method provided in an embodiment of this application;

[0038] Figure 8 This is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0039] Figure 9 This is a schematic diagram of the structure of another communication device provided in the embodiments of this application;

[0040] Figure 10 This is a schematic diagram of the structure of a chip provided in an embodiment of this application;

[0041] Figure 11 This is a schematic diagram of the structure of a module device provided in an embodiment of this application. Detailed Implementation

[0042] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0043] The terms "first" and "second," etc., used in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0044] See Figure 2 This is a schematic diagram of a communication system architecture for information feedback provided in an embodiment of this application. For example... Figure 2 As shown, the communication system includes a network device 201 and multiple terminal devices 202. The network device 201 is equipped with multiple antennas, each of which can communicate with one or more terminal devices 202. When the network device 201 communicates with multiple terminal devices 202 simultaneously, it is necessary to obtain the channel state of the downlink channel corresponding to each terminal device 202. This allows for precoding of the communication information sent to different terminal devices 202 based on the channel state, eliminating interference between the multiple terminal devices 202 and ensuring that each terminal device accurately receives the corresponding communication information.

[0045] The following uses a terminal device as an example to illustrate the process by which terminal device 202 feeds back the channel state of the downlink channel to network device 201. Specifically, terminal device 202 first estimates the downlink channel to obtain a channel matrix (the channel matrix reflects the channel state of the downlink channel); then, based on the segmentation point corresponding to terminal device 202 (exemplary, this segmentation point is related to the specific performance of terminal device 202 and the current transmission environment), it determines a first processing model and uses the first processing model to compress the channel matrix to obtain a compressed channel matrix; the compressed channel matrix and the segmentation point identifier together serve as channel state information, which is sent by terminal device 202 to network device 201. After receiving the channel state information, network device 201 first determines a second processing model based on the segmentation point identifier, and then uses the second processing model to decompress the compressed channel matrix in the channel state information to obtain the channel matrix. Based on the decompressed channel matrix, network device 201 can determine the coding and modulation scheme, physical resource allocation scheme, and other configurations used for downlink communication with terminal device 202. In this process, the data size (or dimension) of the compressed channel matrix is ​​much smaller than the data size of the channel matrix itself, and the data size of the compressed channel matrix changes as the segmentation points change. Therefore, the terminal device in this application can flexibly feed back channel state information to the network device, improving the efficiency of channel state information feedback.

[0046] Optionally, the communication system for the aforementioned information feedback can be a 5th generation (5G) system, also known as an NR system; or it can be applied to a 6th generation (6G) system, a 7th generation (7G) system, or other future communication systems; or it can be used in device-to-device (D2D) systems, machine-to-machine (M2M) systems, long term evolution (LTE) systems, etc. This communication system may include, but is not limited to, one or more network devices (such as network device 201) and one or more terminal devices (such as terminal device 202). Figure 2 The number and form of the devices shown are for illustrative purposes only and do not constitute a limitation on the embodiments of this application.

[0047] In this application embodiment, network device 201 is a device with wireless communication and data processing functions. Network device 201 includes, but is not limited to: evolved node B (eNB), radio network controller (RNC), node B (NB), base station controller (BSC), base transceiver station (BTS), home network device (e.g., home evolved Node B, or home Node B, HNB), baseband unit (BBU), access point (AP), wireless relay node, wireless backhaul node, transmission and reception point (TRP), transmission point (TP) in a wireless fidelity (WIFI) system, etc.; it can also be a device used in 5G, 6G, or even 7G systems, such as gNB in ​​an NR system, or a transmission point (TRP or TP), and this application does not limit this.

[0048] In this embodiment, the terminal device 202 is a device with wireless communication and data processing functions. The terminal device 202 includes, but is not limited to, user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, user agent, or user device. For example, the terminal device may be a mobile phone, tablet computer, computer with wireless transceiver capabilities, virtual reality terminal device, augmented reality terminal device, wireless terminal in industrial control, wireless terminal in autonomous driving, wireless terminal in telemedicine, wireless terminal in smart grids, wireless terminal in transportation safety, wireless terminal in smart cities, wireless terminal in smart homes, wireless terminal in vehicle-to-everything (V2X) communication, etc.

[0049] The following is through Figures 3 to 9 The embodiments shown provide a detailed description of the information feedback method and apparatus proposed in this application.

[0050] See Figure 3 This is a flowchart illustrating an information feedback method provided in an embodiment of this application. This method can be applied to... Figure 2 In the communication system shown, for example, the network device described below can be... Figure 2 Network device 201, terminal device can be Figure 2One of the terminal devices 202 in the process. The method includes steps S301 to S304, wherein:

[0051] S301. The terminal device performs downlink channel estimation to obtain the channel matrix.

[0052] In this process, the terminal device estimates the downlink channel based on the channel state information reference signal (CSI-RS) sent by the network device, obtaining the channel matrix. For example, the network device can insert known CSI-RS into the transmitted data. When the terminal device receives data carrying CSI-RS, it can first obtain the channel estimation result at the location of the CSI-RS, and then use this result to interpolate and obtain the channel estimation result at the data location, which is the channel matrix. The channel matrix reflects the signal fading during downlink channel transmission (i.e., the data carrying CSI-RS). In specific implementations, CSI-RS can be sent periodically or aperiodically by the network device. After receiving any CSI-RS, the terminal device can determine the corresponding channel matrix according to the above process.

[0053] S302. The terminal device compresses the channel matrix based on the first processing sub-model to obtain the compressed channel matrix. The first processing sub-model is determined according to the processing model. The segmentation point corresponding to the terminal device in the processing model and the part before the segmentation point are the first processing sub-model. The part after the segmentation point in the processing model is the second processing sub-model. The second processing sub-model is used to decompress the compressed channel matrix.

[0054] First, the processing model in the embodiments of this application is described: The processing model is a neural network model that processes arbitrary input data to obtain output data that is most similar to the arbitrary input data. This neural network model includes multiple processing layers, which can be used to compress the channel matrix and then restore it to its original form. Exemplarily, this neural network model can be a convolutional neural network (CNN) and its variants, a long short-term memory network (LSTM) and its variants, etc., and this application does not impose any limitations on this.

[0055] It should be noted that, in this embodiment, the processing model is trained as a whole to determine the various parameter settings within it. For example, during training, the input to the processing model is used as its expected output, and the difference between the trained output and the expected output is used as a loss function to update the model, resulting in a trained processing model. Applying this trained model to any input data yields output data with the highest similarity to that input data. Furthermore, within the processing model, the vector dimension of the output data of each processing layer is smaller than the vector dimension of the input data.

[0056] The processing model includes at least one processing layer and at least one segmentation point. A segmentation point is a processing layer within the processing model, and the segmentation point is contained within the at least one segmentation point. For example, the processing layer can be a convolutional layer, a fully connected layer, a pooling layer, etc. The total number of segmentation points included in the at least one segmentation point is less than or equal to the total number of processing layers in the processing model. When the total number of segmentation points included in the at least one segmentation point is equal to the total number of processing layers in the processing model, this application considers each processing layer in the processing model as a segmentation point.

[0057] Specifically, the segmentation point corresponding to the terminal device is one of at least one segmentation point, and the first processing sub-model is composed of the segmentation point corresponding to the terminal device in the processing model and the processing layer before the segmentation point connected in sequence.

[0058] In one possible implementation, the segmentation point and the first processing sub-model corresponding to the terminal device can be determined by the terminal device. For example, the terminal device first determines the segmentation point based on the segmentation point information and first information, and then determines the first processing sub-model based on the segmentation point. The segmentation point information includes one or more of the following: the identifier of at least one segmentation point of the processing model, dimension information, and computational complexity information; the first information includes one or more of the following: the terminal device's capability information, and the channel quality of the channel between the terminal device and the network device; the capability information includes one or more of the following: computing power, communication capability, and power consumption information. The specific implementation process of this method can be referred to... Figure 5 The corresponding embodiments are described below.

[0059] In another possible implementation, the segmentation point and the first processing sub-model corresponding to the terminal device can be indicated by the instruction information sent from the network device to the terminal device. The specific implementation process of this method can be found in [reference needed]. Figure 7 The corresponding embodiments are described below.

[0060] Furthermore, the terminal device compresses the channel matrix according to the first processing sub-model to obtain a compressed channel matrix. In this case, the compressed channel matrix is ​​equivalent to the intermediate output obtained from the segmentation points corresponding to the terminal device in the processing model after inputting the channel matrix. It should be noted that when processing the channel matrix, the first processing sub-model performs dimensionality reduction on the channel matrix. The data dimension of the output data after dimensionality reduction (or intermediate output, compressed channel matrix) is smaller than the data dimension of the channel matrix, which reduces the amount of channel state information that the terminal device needs to transmit.

[0061] S303. The terminal device sends channel state information to the network device. The channel state information includes a compressed channel matrix and the identifier of the segmentation point. The identifier of the segmentation point is used to indicate the segmentation point.

[0062] Since each segmentation point in at least one segmentation point has a corresponding relationship with the identifier of each segmentation point, the identifier of the segmentation point can be used to indicate the segmentation point corresponding to the terminal device.

[0063] Optionally, the compressed channel matrix (i.e., the output of the first processing sub-model) can be concatenated with the identifier of the segmentation point after being quantized into binary bits to form channel state information. The concatenation order can be either the identifier of the segmentation point first, followed by the compressed channel matrix, or vice versa; this application does not impose any restrictions on this. Furthermore, all identifiers of segmentation points (including the identifier of the segmentation point) can be represented by fixed-length binary bits, while the compressed channel matrix, due to the uncertainty in the data size, can be represented by variable-length binary bits.

[0064] S304. After the network device receives the channel state information, the network device decompresses the compressed channel matrix based on the second processing sub-model to obtain the channel matrix; the second processing sub-model is determined based on the segmentation point.

[0065] The part following the segmentation point in the processing model constitutes the second processing sub-model. Specifically, the processing layer following the segmentation point in the processing model is the second processing sub-model. The second processing sub-model is used to decompress the compressed channel matrix.

[0066] In one possible implementation, the network device first determines the segmentation point corresponding to the terminal device based on the identifier of the segmentation point corresponding to the terminal device in the channel state information and the correspondence between the identifier and the segmentation point; then, it determines the second processing sub-model based on the segmentation point corresponding to the terminal device and the processing model; finally, it decompresses the compressed channel matrix based on the second processing sub-model to obtain the channel matrix.

[0067] In one possible implementation, the segmentation point indicated by the segmentation point identifier is used to determine the size of the compressed channel matrix. Specifically, after receiving the channel state information, the network device can first determine the size of the compressed channel matrix (or dimension, with a correspondence between the segmentation point identifier and the dimension information) based on the segmentation point identifier corresponding to the terminal device. Then, the network device performs a completeness check on the compressed channel matrix in the channel state information based on the size of the compressed channel matrix. If the completeness check passes, it means that the network device has received the complete compressed channel matrix. Conversely, if the completeness check fails, the network device can either wait and receive the unreceived portion of the compressed channel matrix, or send a prompt message to the terminal device, requesting the terminal device to resend the channel state information.

[0068] Furthermore, network devices can determine the coding and modulation scheme, physical resource allocation scheme, and other configurations used when communicating downlink with terminal devices based on the channel matrix.

[0069] The following is based on Figure 4 For example, let's illustrate the information feedback methods mentioned above. Figure 4 As shown, the processing model consists of seven processing layers connected sequentially: one convolutional layer, two fully connected layers, and four convolutional layers. When the segmentation point corresponding to the terminal device is the first fully connected layer, the first processing sub-model determined by this segmentation point is composed of the first two processing layers of the processing model connected sequentially, and the second processing sub-model is composed of the last five processing layers of the processing model connected sequentially.

[0070] On the terminal device side, the terminal device first obtains the channel matrix based on the downlink channel estimation, and then calls... Figure 4 The first processing sub-model in the model compresses the channel matrix, and the output of the fully connected layer in the first processing sub-model is the compressed channel matrix (i.e., Figure 4 The intermediate output is then quantized into binary bits and concatenated with the identifier of the segmentation point corresponding to the terminal device (which can be represented by binary bits) to obtain the channel state information (i.e., the intermediate output). Figure 4 The terminal device sends the channel state information to the network device (using the code "001011" in the code); finally, the terminal device sends the channel state information to the network device. On the network device side, after receiving the channel state information, the network device calls... Figure 4 The second processing sub-model in the model processes the compressed channel matrix (i.e., ...) from the channel state information. Figure 4 The intermediate output is decompressed to obtain the channel matrix.

[0071] exist Figure 3In a corresponding embodiment, this application can divide the processing model into a first processing model and a second processing model based on a segmentation point corresponding to the terminal device. The terminal device uses the first processing model to compress the channel matrix, and the network device uses the second processing model to recover the compressed channel matrix. During this process, the selection of the segmentation point is related to the terminal device, and the first and second processing models change with the segmentation point. The dimension (or size) of the compressed channel matrix obtained through the first processing model also changes accordingly. Therefore, in this application, the terminal device can flexibly feed back channel state information to the network device, improving the efficiency of channel state information feedback.

[0072] The above Figure 3 The corresponding implementation example introduces the overall process of information feedback, which will be described below. Figures 5-7 The illustrated embodiment describes how to determine the segmentation point and the first processing sub-model corresponding to the terminal device. Among them, Figure 5 In the corresponding embodiment, the segmentation point and the first processing sub-model are determined by the terminal device. Figure 7 In the corresponding embodiment, the segmentation point and the first processing sub-model are determined by the network device.

[0073] See Figure 5 This is a flowchart illustrating another information feedback method provided in an embodiment of this application. This method can be applied to... Figure 2 In the communication system shown, for example, the network device described below can be... Figure 2 Network device 201, terminal device can be Figure 2 One of the terminal devices 202 in the process. The method includes steps S501 to S509, wherein:

[0074] S501, Network device determines processing model, at least one segmentation point and segmentation point information of at least one segmentation point.

[0075] This application allows the processing model to be trained in a network device to determine various parameter settings within the processing model. Alternatively, the network device in this application can directly receive a trained processing model from other devices. The trained processing model includes various predetermined parameter settings. For example, these parameters include, but are not limited to, the number of neurons in each processing layer of the processing model, activation function settings, weight settings, and input and output data dimension settings, etc.

[0076] Furthermore, network devices can parse the processing model to obtain the dimensionality and computational complexity information of each processing layer within the model. The dimensionality information of a processing layer indicates the dimension of the data output by that layer, while the computational complexity information indicates the computational complexity required for the corresponding sub-model to run (this computational complexity is the number of floating-point operations; a multiplication or addition counts as one operation). The corresponding sub-model is the processing layer and all preceding processing layers within the processing model. Specifically, the sub-model corresponding to a processing layer includes both the current processing layer and all preceding processing layers within the processing model, and the computational complexity information of a processing layer is the sum of the number of floating-point operations performed by the current processing layer and all preceding processing layers within the processing model.

[0077] Furthermore, the network device can determine at least one segmentation point based on the processing model. In one possible implementation, the network device can use each processing layer in the processing model as one of the at least one segmentation points, and the total number of segmentation points included in the at least one segmentation point is equal to the total number of processing layers included in the processing model.

[0078] In another possible implementation, the network device first needs to determine the maximum number of at least one segmentation points, and then select processing layers from the processing model whose total number is less than or equal to the maximum value, using these selected processing layers as the segmentation points among the at least one segmentation points. The maximum number of at least one segmentation points is determined by the number of bits occupied by the segmentation point's identifier. The number of bits occupied by the identifier is the maximum number of bits occupied by the identifier considering transmission overhead. If the maximum number of at least one segmentation points is N, and the maximum number of bits occupied by the identifier is M, then the relationship between N and M can be expressed as N = 2^N. M express.

[0079] For example, in order to minimize transmission overhead, if the maximum number of bits M occupied by the identifier of the segmentation point that the current channel can support is 2, then the maximum number N of at least one segmentation point is 4.

[0080] Furthermore, the network device can determine at least one segmentation point based on the maximum number of determined at least one segmentation points and the dimensional and computational information of each processing layer in the processing model. Optionally, to ensure the distinguishability of the computational and dimensional information of each segmentation point, this application can use computational difference thresholds and dimensional difference thresholds to limit the minimum difference in computational and dimensional information between any two adjacent segmentation points. Moreover, the network device can also limit the maximum value of the computational and dimensional information of the segmentation point based on the capabilities of the terminal device.

[0081] For example, with Figure 6 For example, the figure shows a bar chart illustrating the relevant information of each processing layer in a processing model provided in this application embodiment. The processing model includes seven processing layers. The white bars represent the single-layer floating-point calculation volume of each processing layer in the processing model, and the black bars represent the dimension of the output data of each processing layer. The dimension information of each processing layer is 2.0, 1.5, 1.3, 0.4, 0.4, 0.8, and 0.8, respectively. The calculation volume information of each processing layer (the calculation volume information is the sum of the single-layer floating-point calculation volumes of each processing layer in the corresponding processing sub-model) is 0, 28, 36, 44, 52, 82, and 89, respectively. For example, when the dimension difference threshold is 0.2, the calculation volume difference threshold is 8, the maximum dimension value is 1.8, and the maximum calculation volume value is 80, the first processing layer is not considered because its dimension information is too large (for example, 2.0 is greater than 1.8). The second and third processing layers can be selected as splitting points. The fourth and fifth processing layers are not suitable to be selected together as split points (the difference in dimensions between these two layers is 0, less than 0.2, indicating that the processing sub-models determined by both layers do not significantly change the dimensions of the output information). The sixth and seventh processing layers are also unsuitable as split points (the computational cost of these two layers is greater than the maximum computational cost of 80, indicating that the processing sub-models determined by both layers place excessive demands on the capabilities of the terminal device). While the fourth and fifth processing layers are not suitable to be selected together as split points, since their dimensional information is the same (and the difference in dimensional information between them and the third layer is 0.9, exceeding 0.2), and their computational cost is not significantly different, either layer can be arbitrarily selected as a split point. For example, Figure 6 The network device selects the 5th processing layer as the split point. Based on this, the network device can select the 2nd, 3rd, and 5th processing layers from the 7 processing layers as at least one split point: split point 0, split point 1, and split point 2. It should be noted that the above selection process is only an example, and in specific applications, various constraints (such as computational difference threshold, maximum computational value, etc.) can be adjusted according to the actual situation.

[0082] S502, The network device sends the processing model and at least one segmentation point information to the terminal device.

[0083] After the network device determines the processing model, the dimensional information and computational complexity information of each processing layer in the processing model, and at least one segmentation point through step S501, it can further determine the identifier of each segmentation point among the at least one segmentation point. The identifier and the segmentation point have a corresponding relationship. For example, Table 1 is a correspondence table between segmentation points and their identifiers provided in an embodiment of this application. The segmentation points included in this correspondence table are... Figure 6 The corresponding example shows the three segmentation points determined by the network device.

[0084] Table 1

[0085] Dividing point Dividing point markings Second processing layer 00 3rd processing layer 01 5th processing layer 10

[0086] As shown in Table 1, when the identifier of the network device segmentation point is represented by two binary bits, the identifier of the second processing layer can be represented by "00", the third processing layer can be represented by "01", and the fifth processing layer can be represented by "10".

[0087] Furthermore, the network device sends the processing model and segmentation point information to the terminal device. In one possible implementation, the segmentation point information includes the identifier of at least one segmentation point. In another possible implementation, the segmentation point information includes the identifier of at least one segmentation point, dimension information, and computational complexity information.

[0088] S503, The terminal device performs downlink channel estimation to obtain the channel matrix.

[0089] After the terminal device receives the processing model (equivalent to receiving various parameter settings of the processing model) and the segmentation point information of at least one segmentation point sent by the network device, if the terminal device receives the Channel State Information Reference Signal (CSI-RS), it can perform downlink channel estimation based on the CSI-RS to obtain the channel matrix. For details, please refer to [link to implementation details]. Figure 3 The description in step S301 in the corresponding embodiment.

[0090] S504. The terminal device determines the segmentation point corresponding to the terminal device based on the processing model, the segmentation point information of at least one segmentation point, and the first information.

[0091] Specifically, the terminal device determines the segmentation point corresponding to the terminal device based on segmentation point information of at least one segmentation point and first information. The first information includes one or more of the following: terminal device capability information, channel quality of the channel between the terminal device and the network device; the capability information includes one or more of the following: computing power, communication power, and power information. Specifically, computing power is used to indicate the number of floating-point operations performed by the terminal device in the processing layer per unit time, communication power is used to indicate the data transmission rate of the terminal device, and power information is used to indicate the remaining storage energy of the terminal device.

[0092] In one possible implementation, if the segmentation point information received by the terminal device includes the identifier of at least one segmentation point, the specific process by which the terminal device determines the segmentation point corresponding to the terminal device based on the segmentation point information and the first information is as follows: The terminal device determines which processing layers in the processing model are segmentation points among the at least one segmentation point based on the identifier of the at least one segmentation point. Next, the terminal device parses the processing model to determine the dimension information and computational complexity information corresponding to each segmentation point among the at least one segmentation point. Finally, the terminal device selects one segmentation point from the at least one segmentation point as the segmentation point corresponding to the terminal device based on the parsed dimension information, computational complexity information, and the aforementioned first information.

[0093] On the one hand, the terminal device determines the time required to run the processing sub-model and feedback channel status information based on computational load information, dimensional information, computing power and communication capabilities. This time can be compared with the terminal device's power information to determine whether the terminal device has enough energy to complete the operation of feedback channel status information.

[0094] For example, as described above Figure 6 For example, if the terminal device's computing power is 200 floating-point calculations per second, then the time required for the terminal device to run the processing sub-model corresponding to segment point 0 is 28 / 200 = 140ms, the time required to run the processing sub-model corresponding to segment point 1 is 36 / 200 = 180ms, and the time required to run the processing sub-model corresponding to segment point 2 is 52 / 200 = 260ms. If the terminal device's communication capability is 20MB of data to be sent per second, then the time required for the terminal device to send the output data of segment point 0 is 1.5 / 20 = 75ms, the time required to send the output data of segment point 1 is 1.3 / 20 = 65ms, and the time required to send the output data of segment point 2 is 0.4 / 20 = 20ms. Furthermore, for segment points 0, 1, and 2, the total time required by the terminal device is 140 + 75 = 215ms, 180 + 65 = 245ms, and 260 + 20 = 285ms, respectively. The total time required by the terminal device is compared with the terminal device's power information (remaining stored energy), and the segmentation point that meets the power information condition is selected.

[0095] On the other hand, the terminal device determines whether the current channel can effectively transmit the data output by the processing sub-model corresponding to the segmentation point based on the dimensional information and the channel quality between the terminal device and the network device.

[0096] For example, the channel quality between the terminal device and the network device can be characterized by the maximum amount of data that the channel can transmit at the current moment. The maximum amount of data is compared with the dimension of the output data of the processing sub-model corresponding to the segmentation point, and the segmentation point corresponding to the dimension of the output data that is less than the maximum amount of data is selected.

[0097] It should be noted that the terminal device may use a segmentation point that satisfies both of the above-mentioned constraints as its corresponding segmentation point; or it may use a segmentation point that satisfies either of the above-mentioned constraints as its corresponding segmentation point. This application does not impose any restrictions on this. For example, if the channel quality is good, the final corresponding segmentation point can be determined based on the total time required by the terminal device and the terminal device's power information. If the terminal device currently has a large amount of stored energy and is in an idle state, the final corresponding segmentation point can be determined based on the maximum amount of data transmitted through the channel and the dimension of the output data of the processing sub-model corresponding to the segmentation point.

[0098] In another possible implementation, if the segmentation point information received by the terminal device includes the identifier of at least one segmentation point, dimension information, and computational complexity information, then the specific process by which the terminal device determines the segmentation point corresponding to the terminal device based on the segmentation point information and the first information is as follows: The terminal device determines which processing layers in the processing model are segmentation points among the at least one segmentation point based on the identifier of at least one segmentation point. Then, the terminal device selects one segmentation point from the at least one segmentation point as the segmentation point corresponding to the terminal device based on the dimension information, computational complexity information, and the first information (the selection process can be found in the above possible implementations and will not be repeated here).

[0099] In this possible implementation, the terminal device does not need to determine the dimension information and computational complexity of at least one segmentation point through parsing the processing model, thus improving the processing efficiency of the terminal device.

[0100] S505. The terminal device determines the first processing sub-model based on the segmentation point corresponding to the terminal device.

[0101] After the terminal device determines the segmentation point corresponding to the terminal device according to the possible implementation in step S504, the model formed by sequentially connecting the segmentation point and the processing layer before the segmentation point in the processing model can be used as the first processing sub-model.

[0102] S506. The terminal device compresses the channel matrix based on the first processing sub-model to obtain the compressed channel matrix.

[0103] The compressed channel matrix is ​​equivalent to the intermediate output obtained from the segmentation points corresponding to the terminal devices in the processing model after the channel matrix is ​​input into the processing model. The data dimension of the compressed channel matrix (i.e., the dimension information of the segmentation points corresponding to the terminal devices) is smaller than the data dimension of the channel matrix.

[0104] S507. The terminal device sends channel state information to the network device. The channel state information includes the compressed channel matrix and the identifier of the segmentation point corresponding to the terminal device.

[0105] S508, the network device determines the second processing sub-model based on the identifier of the segmentation point corresponding to the terminal device.

[0106] S509. The network device decompresses the compressed channel matrix based on the second processing sub-model to obtain the channel matrix.

[0107] For detailed implementation methods of steps S507 to S509, please refer to [link to relevant documentation]. Figure 3 The corresponding descriptions in steps S303 to S304 of the corresponding embodiments will not be repeated here.

[0108] exist Figure 5 In a corresponding embodiment, the network device can determine at least one segmentation point based on the dimensional information and computational information of each processing layer in the processing model, and the terminal device can determine the segmentation point corresponding to the terminal device based on the segmentation point information of at least one segmentation point and the first information. The segmentation point can be used to determine the first processing sub-model so that the terminal device can obtain the channel state information through the first processing sub-model and feed the channel state information back to the network device.

[0109] See Figure 7 This is a flowchart illustrating another information feedback method provided in an embodiment of this application. This method can be applied to... Figure 2 In the communication system shown, for example, the network device described below can be... Figure 2 Network device 201, terminal device can be Figure 2 One of the terminal devices 202 in the process. The method includes steps S7001 to S7010, wherein:

[0110] S7001, Network device determines processing model, at least one segmentation point and segmentation point information of at least one segmentation point.

[0111] In this process, the network device first determines the processing model, and then determines at least one segmentation point based on the dimensional information and computational information of each processing layer in the processing model. The segmentation point information of at least one segmentation point includes the identifier of at least one segmentation point, or includes the identifier of at least one segmentation point, dimensional information and / or computational information.

[0112] In this embodiment, step S7001 and Figure 5 The implementation of step S501 is the same in the corresponding embodiments, and will not be repeated here.

[0113] S7002, The terminal device sends the first information to the network device.

[0114] Optionally, the network device may send a request to the terminal device to request first information related to the terminal device. The first information includes one or more of the following: capability information of the terminal device, channel quality of the channel between the terminal device and the network device; the capability information includes one or more of the following: computing power, communication power, and power information.

[0115] S7003, the network device determines the segmentation point corresponding to the terminal device based on the processing model, the dimension information and computational quantity information of at least one segmentation point, and the first information.

[0116] After receiving the first information sent by the terminal device, the network device will select a corresponding segmentation point for the terminal device based on the dimension information and computational complexity information of at least one segmentation point determined in step S7001 and the first information. The network device can select the corresponding segmentation point for the terminal device based on the following two aspects.

[0117] On the one hand, the network device determines the time required for the terminal device to run the processing sub-model of the segmentation point and the feedback channel status information based on computational load information, dimensional information, computing power, and communication capabilities. This time can be compared with the terminal device's power information to determine whether the terminal device has sufficient energy to complete the operation of feeding back the channel status information. On the other hand, the network device determines whether the current channel can effectively transmit the data output by the processing sub-model of the segmentation point based on dimensional information and the channel quality between the terminal device and the network device. It should be noted that the network device can use a segmentation point that simultaneously satisfies the above two aspects as the segmentation point corresponding to the terminal device; or it can use a segmentation point that satisfies the constraints of either of the above two aspects as the segmentation point corresponding to the terminal device. This application does not impose any restrictions on this, and specific implementation methods for these two aspects can be found in [reference needed]. Figure 5 The corresponding description in S504 of the corresponding embodiment will not be repeated here.

[0118] S7004. The network device determines the first processing sub-model based on the segmentation point corresponding to the terminal device.

[0119] After the network device determines the segmentation point corresponding to the terminal device according to the implementation method in step S7003, the model formed by sequentially connecting the segmentation point and the processing layer before the segmentation point in the processing model can be used as the first processing sub-model.

[0120] S7005. The network device sends the identifier of the segmentation point corresponding to the terminal device and the first processing sub-model to the terminal device.

[0121] Based on step S7005, the network device only needs to send the corresponding segmentation point identifier and the first processing sub-model to the terminal device, without sending the complete processing model, the identifiers of all segmentation points in at least one segmentation point, and other segmentation point information (such as dimension information and computational quantity information). This can reduce the amount of data sent by the network device to the terminal device and improve the processing efficiency of the terminal device.

[0122] S7006. The terminal equipment performs downlink channel estimation to obtain the channel matrix.

[0123] S7007. The terminal device compresses the channel matrix based on the first processing sub-model to obtain the compressed channel matrix.

[0124] S7008. The terminal device sends channel state information to the network device. The channel state information includes the compressed channel matrix and the identifier of the segmentation point corresponding to the terminal device.

[0125] S7009. The network device determines the second processing sub-model based on the identifier of the segmentation point corresponding to the terminal device.

[0126] S7010, the network device decompresses the compressed channel matrix based on the second processing sub-model to obtain the channel matrix.

[0127] The implementation methods for steps S7006 to S7010 can be found in [reference needed]. Figure 5 The corresponding descriptions in steps S503, S506 to S509 in the corresponding embodiments will not be repeated here.

[0128] exist Figure 7 In a corresponding embodiment, the network device can determine at least one segmentation point based on the dimensional information and computational complexity information of each processing layer in the processing model. Based on the dimensional information, computational complexity information, and first information of the at least one segmentation point, the network device can determine the segmentation point and the first processing sub-model corresponding to the terminal device. Then, the network device can send the identifier of the segmentation point corresponding to the terminal device and the first processing sub-model to the terminal device, so that the terminal device can obtain the channel state information through the first processing sub-model and feed back the channel state information to the network device. This method reduces the processing work on the terminal device side and improves the efficiency of the terminal device in feeding back the channel state information.

[0129] It is understood that, in order to implement the functions in the above embodiments, the terminal device and network device include hardware structures and / or software modules corresponding to perform each function. Those skilled in the art should readily recognize that, based on the units and method steps of the various examples described in conjunction with the embodiments disclosed in this application, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application scenario and design constraints of the technical solution.

[0130] See Figure 8 This is a schematic diagram of a communication device provided in an embodiment of this application. The communication device 80 includes a processing unit 801, a transmitting unit 802, and a receiving unit 803.

[0131] In one embodiment, the communication device 80 may be a terminal device or a device for a terminal device. The device for the terminal device may be a chip system or a chip within a user equipment. The chip system may consist of chips or may include chips and other discrete components.

[0132] Processing unit 801 is used to perform downlink channel estimation to obtain the channel matrix;

[0133] The processing unit 801 is further configured to compress the channel matrix based on a first processing sub-model to obtain a compressed channel matrix; wherein, the first processing sub-model is determined according to a processing model, the segmentation point corresponding to the terminal device in the processing model and the part before the segmentation point are the first processing sub-model, the part after the segmentation point in the processing model are the second processing sub-model, and the second processing sub-model is used to decompress the compressed channel matrix.

[0134] The transmitting unit 802 is used to transmit channel state information to the network device. The channel state information includes the compressed channel matrix and the identifier of the segmentation point, and the identifier of the segmentation point is used to indicate the segmentation point.

[0135] In one possible implementation, the segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

[0136] In one possible implementation, the processing model includes at least one processing layer, the processing model includes at least one segmentation point, a segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point;

[0137] The segmentation point corresponding to the terminal device in the processing model and the portion before the segmentation point constitute the first processing sub-model, including:

[0138] The segmentation point corresponding to the terminal device in the processing model and the processing layer before the segmentation point constitute the first processing sub-model.

[0139] In one possible implementation, the segmentation point is determined based on segmentation point information and first information, wherein the segmentation point information includes the identifier of at least one segmentation point of the processing model; the first information includes one or more of the following: capability information of the terminal device, channel quality of the channel between the terminal device and the network device; the capability information includes one or more of the following: computing power, communication power, and power information.

[0140] In one possible implementation, the segmentation point information further includes the dimension information and computational complexity information of the at least one segmentation point; the dimension information of a segmentation point is used to indicate the dimension of the data output by the segmentation point, and the computational complexity information of a segmentation point is used to indicate the computational complexity required for the processing sub-model corresponding to the segmentation point to run; the processing sub-model corresponding to a segmentation point is the segmentation point and the part before the segmentation point in the processing model.

[0141] In one possible implementation, the receiving unit 803 is configured to receive the segmentation point information from the network device.

[0142] In one possible implementation, the dimensional information and computational complexity information of the at least one segmentation point are determined based on the identifier of the at least one segmentation point and the processing model.

[0143] In one possible implementation, the receiving unit 803 is further configured to receive the processing model from the network device.

[0144] Specifically, Figure 8 The operations performed by each unit of the communication device 80 shown can be referred to the above. Figures 3-7 The corresponding method embodiments include details regarding the terminal device, which will not be elaborated here. The aforementioned units can be implemented in hardware, software, or a combination of both. In one embodiment, the functions of each unit described above can be implemented by one or more processors in the communication device 80.

[0145] In another embodiment, the communication device 80 can be a network device or a means for a network device. The means for a network device can be a chip system or a chip within the network device. The chip system can be composed of chips or can include chips and other discrete components.

[0146] The receiving unit 803 is used to receive channel state information, which includes a channel matrix compressed using a first processing sub-model and identifiers of segmentation points corresponding to the terminal device; wherein, the first processing sub-model is determined according to a processing model, the identifiers of the segmentation points are used to indicate the segmentation points, the segmentation points and the portion before the segmentation points in the processing model constitute the first processing sub-model, the portion after the segmentation points in the processing model constitutes the second processing sub-model, and the second processing model is used to decompress the compressed channel matrix;

[0147] Processing unit 801 is used to decompress the compressed channel matrix based on the second processing sub-model to obtain the channel matrix; the second processing sub-model is determined according to the segmentation point.

[0148] In one possible implementation, the segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

[0149] In one possible implementation, the processing model includes at least one processing layer, the processing model includes at least one segmentation point, a segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point;

[0150] The portion following the segmentation point in the processing model constitutes the second processing sub-model, which includes:

[0151] The processing layer following the segmentation point in the processing model is the second processing sub-model.

[0152] In one possible implementation, the at least one segmentation point is determined based on the dimensional information and computational complexity information of each processing layer in the processing model; the dimensional information of a processing layer is used to indicate the dimension of the data output by the processing layer, and the computational complexity information of a processing layer is used to indicate the computational complexity required to run the processing sub-model corresponding to the processing layer; the processing sub-model corresponding to a processing layer is the processing layer and the part before the processing layer in the processing model.

[0153] In one possible implementation, the maximum number of the at least one segmentation point is determined by the number of bits occupied by the identifier of the segmentation point.

[0154] In one possible implementation, the sending unit 802 is configured to send the processing model and segmentation point information to the terminal device, wherein the segmentation point information includes the identifier of the at least one segmentation point.

[0155] In one possible implementation, the segmentation point information further includes the dimension information and computational cost information corresponding to the at least one segmentation point.

[0156] It should be noted that the functions of each unit module of the communication device in the embodiments of this application can be specifically implemented according to the methods in the above method embodiments. The specific implementation process can be referred to the relevant descriptions in the above method embodiments, which will not be repeated here.

[0157] Specifically, Figure 8 The operations performed by each unit of the communication device 80 shown can be referred to the above. Figures 3-7 The corresponding method embodiments contain details related to network devices, which will not be elaborated here. The aforementioned units can be implemented in hardware, software, or a combination of both. In one embodiment, the functions of each unit described above can be implemented by one or more processors in the communication device 80.

[0158] See Figure 9 This is a schematic diagram of another communication device provided in an embodiment of this application, used to implement the above. Figures 3-7 The functions of a terminal device or network device. In one embodiment, the communication device 90 is used to implement the above. Figures 3-7 The communication device can be a terminal device or a device for a terminal device. The device for a terminal device can be a chip system or a chip within a user equipment. The chip system can consist of chips or include chips and other discrete components. In another embodiment, the communication device 90 is used to implement the above-described functionality. Figures 3-7 The function of the network device in the communication device. This communication device can be a network device or a device for a network device. The device for a network device can be a chip system or a chip within the network device. The chip system can consist of chips or may include chips and other discrete components.

[0159] The communication device 90 includes at least one processor 902 for implementing the data processing functions of the terminal device / network device in the method provided in this application embodiment. The communication device 90 may also include a communication interface 901 for implementing the transmit and receive operations of the terminal device / network device in the method provided in this application embodiment. In this application embodiment, the communication interface may be a transceiver, circuit, bus, module, or other type of communication interface for communicating with other devices via a transmission medium. For example, the communication interface 901 enables the device in the communication device 90 to communicate with other devices. The processor 902 uses the communication interface 901 to transmit and receive data and is used to implement the above method embodiments. Figures 3-7 The method described.

[0160] The communication device 90 may further include at least one memory 903 for storing program instructions and / or data. The memory 903 is coupled to the processor 902. The coupling in this embodiment is an indirect coupling or communication connection between devices, units, or modules, and may be electrical, mechanical, or other forms, used for information exchange between devices, units, or modules. The processor 902 may operate in conjunction with the memory 903. The processor 902 may execute program instructions stored in the memory 903. At least one of the at least one memories may be included in the processor.

[0161] When the communication device 90 is powered on, the processor 902 can read the software program in the memory 903, interpret and execute the instructions of the software program, and process the data of the software program. When data needs to be transmitted wirelessly, the processor 902 performs baseband processing on the data to be transmitted and outputs the baseband signal to the radio frequency circuit (not shown in the figure). The radio frequency circuit processes the baseband signal and transmits the radio frequency signal outward in the form of electromagnetic waves through the antenna. When data is sent to the communication device 90, the radio frequency circuit receives the radio frequency signal through the antenna, converts the radio frequency signal into a baseband signal, and outputs the baseband signal to the processor 902. The processor 902 converts the baseband signal into data and processes the data.

[0162] In another implementation, the radio frequency circuit and antenna can be set up independently of the processor 902 that performs baseband processing. For example, in a distributed scenario, the radio frequency circuit and antenna can be arranged in a remote manner, independent of the communication device.

[0163] This application embodiment does not limit the specific connection medium between the communication interface 901, processor 902, and memory 903. This application embodiment... Figure 9 The memory 903, processor 902, and communication interface 901 are connected via a bus 904. Figure 9 The connections between other components are shown in bold and are for illustrative purposes only, not as limiting information. The bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0164] When the communication device 90 is specifically used for a terminal device / network device, for example, when the communication device 90 is specifically a chip or chip system, the communication interface 901 can output or receive baseband signals. When the communication device 90 is specifically a terminal device / network device, the communication interface 901 can output or receive radio frequency signals. In the embodiments of this application, the processor can be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, which can implement or execute the various methods, operations, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor, etc. The operation of the method disclosed in the embodiments of this application can be directly reflected as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0165] It should be noted that the communication device can perform the relevant steps of the terminal device / network device in the aforementioned method embodiments. For details, please refer to the implementation methods provided in the above steps, which will not be repeated here.

[0166] For various devices and products applied to or integrated into communication devices, each of its modules can be implemented using hardware such as circuits. Different modules can be located in the same component (e.g., chip, circuit module, etc.) or different components within the terminal. Alternatively, at least some modules can be implemented using software programs that run on a processor integrated within the terminal, while the remaining (if any) modules can be implemented using hardware such as circuits.

[0167] For cases where the communication device can be a chip or a chip system, please refer to [link / reference]. Figure 10 The diagram shows the structure of the chip. The chip 100 includes a processor 1001 and a communication interface 1002. The number of processors 1001 can be one or more, and the number of communication interfaces 1002 can be multiple.

[0168] In one embodiment, the processor 1001 is configured to perform the following operations:

[0169] Downlink channel estimation is performed to obtain a channel matrix; the channel matrix is ​​compressed based on a first processing sub-model to obtain a compressed channel matrix; wherein, the first processing sub-model is determined according to a processing model, the segmentation point corresponding to the terminal device in the processing model and the part before the segmentation point are the first processing sub-model, and the part after the segmentation point in the processing model are the second processing sub-model, the second processing sub-model is used to decompress the compressed channel matrix; channel state information is sent to the network device, the channel state information includes the compressed channel matrix and the identifier of the segmentation point, the identifier of the segmentation point is used to indicate the segmentation point.

[0170] In one possible implementation, the segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

[0171] In one possible implementation, the processing model includes at least one processing layer, the processing model includes at least one segmentation point, a segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point; the segmentation point corresponding to the terminal device in the processing model and the portion before the segmentation point constitute the first processing sub-model, including: the segmentation point corresponding to the terminal device in the processing model and the processing layer before the segmentation point constitute the first processing sub-model.

[0172] In one possible implementation, the segmentation point is determined based on segmentation point information and first information, wherein the segmentation point information includes the identifier of at least one segmentation point of the processing model; the first information includes one or more of the following: capability information of the terminal device, channel quality of the channel between the terminal device and the network device; the capability information includes one or more of the following: computing power, communication power, and power information.

[0173] In one possible implementation, the segmentation point information further includes the dimension information and computational complexity information of the at least one segmentation point; the dimension information of a segmentation point is used to indicate the dimension of the data output by the segmentation point, and the computational complexity information of a segmentation point is used to indicate the computational complexity required for the processing sub-model corresponding to the segmentation point to run; the processing sub-model corresponding to a segmentation point is the segmentation point and the part before the segmentation point in the processing model.

[0174] In one possible implementation, the processor 1001 is also configured to perform the following operation: receiving the segmentation point information from the network device.

[0175] In one possible implementation, the dimensional information and computational complexity information of the at least one segmentation point are determined based on the identifier of the at least one segmentation point and the processing model.

[0176] In one possible implementation, the processor 1001 is also configured to perform the following operation: receiving the processing model from the network device.

[0177] In another embodiment, the processor 1001 is configured to perform the following operations:

[0178] The system receives channel state information, which includes a channel matrix compressed using a first processing sub-model and identifiers of segmentation points corresponding to the terminal device. The first processing sub-model is determined based on a processing model. The identifiers of the segmentation points indicate the segmentation points. The segmentation point and the portion preceding it in the processing model constitute the first processing sub-model, while the portion following it constitutes a second processing sub-model. The second processing sub-model is used to decompress the compressed channel matrix. The compressed channel matrix is ​​then decompressed based on the second processing sub-model to obtain the channel matrix. The second processing sub-model is determined based on the segmentation points.

[0179] In one possible implementation, the segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

[0180] In one possible implementation, the processing model includes at least one processing layer, the processing model includes at least one segmentation point, a segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point; the portion of the processing model after the segmentation point is a second processing sub-model, including: the processing layer after the segmentation point in the processing model is the second processing sub-model.

[0181] In one possible implementation, the at least one segmentation point is determined based on the dimensional information and computational complexity information of each processing layer in the processing model; the dimensional information of a processing layer is used to indicate the dimension of the data output by the processing layer, and the computational complexity information of a processing layer is used to indicate the computational complexity required to run the processing sub-model corresponding to the processing layer; the processing sub-model corresponding to a processing layer is the processing layer and the part before the processing layer in the processing model.

[0182] In one possible implementation, the maximum number of the at least one segmentation point is determined by the number of bits occupied by the identifier of the segmentation point.

[0183] In one possible implementation, the processor 1001 is further configured to perform the following operation: sending the processing model and segmentation point information to the terminal device, the segmentation point information including the identifier of the at least one segmentation point.

[0184] In one possible implementation, the segmentation point information further includes the dimension information and computational cost information corresponding to the at least one segmentation point.

[0185] For each device or product applied to or integrated into the chip, each of its modules can be implemented using hardware methods such as circuits, or at least some modules can be implemented using software programs that run on the processor 1001 integrated inside the chip, and the remaining (if any) modules can be implemented using hardware methods such as circuits.

[0186] See Figure 11 This is a schematic diagram of the structure of a module device provided in an embodiment of this application. The module device 110 can perform the relevant steps of the first terminal in the aforementioned method embodiment. The module device 110 includes: a communication module 1101, a power module 1102, a storage module 1103, and a chip module 1104.

[0187] The power module 1102 is used to provide power to the module device; the storage module 1103 is used to store data and instructions; and the communication module 1101 is used for internal communication within the module device or for communication between the module device and external devices.

[0188] In one embodiment, the chip module 1104 is used for:

[0189] Downlink channel estimation is performed to obtain a channel matrix; the channel matrix is ​​compressed based on a first processing sub-model to obtain a compressed channel matrix; wherein, the first processing sub-model is determined according to a processing model, the segmentation point corresponding to the terminal device in the processing model and the part before the segmentation point are the first processing sub-model, and the part after the segmentation point in the processing model are the second processing sub-model, the second processing sub-model is used to decompress the compressed channel matrix; channel state information is sent to the network device, the channel state information includes the compressed channel matrix and the identifier of the segmentation point, the identifier of the segmentation point is used to indicate the segmentation point.

[0190] In one possible implementation, the segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

[0191] In one possible implementation, the processing model includes at least one processing layer, the processing model includes at least one segmentation point, a segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point; the segmentation point corresponding to the terminal device in the processing model and the portion before the segmentation point constitute the first processing sub-model, including: the segmentation point corresponding to the terminal device in the processing model and the processing layer before the segmentation point constitute the first processing sub-model.

[0192] In one possible implementation, the segmentation point is determined based on segmentation point information and first information, wherein the segmentation point information includes the identifier of at least one segmentation point of the processing model; the first information includes one or more of the following: capability information of the terminal device, channel quality of the channel between the terminal device and the network device; the capability information includes one or more of the following: computing power, communication power, and power information.

[0193] In one possible implementation, the segmentation point information further includes the dimension information and computational complexity information of the at least one segmentation point; the dimension information of a segmentation point is used to indicate the dimension of the data output by the segmentation point, and the computational complexity information of a segmentation point is used to indicate the computational complexity required for the processing sub-model corresponding to the segmentation point to run; the processing sub-model corresponding to a segmentation point is the segmentation point and the part before the segmentation point in the processing model.

[0194] In one possible implementation, the chip module 1104 is further configured to: receive the segmentation point information from the network device.

[0195] In one possible implementation, the dimensional information and computational complexity information of the at least one segmentation point are determined based on the identifier of the at least one segmentation point and the processing model.

[0196] In one possible implementation, the chip module 1104 is further configured to: receive the processing model from the network device.

[0197] In another embodiment, the chip module 1104 is used for:

[0198] The system receives channel state information, which includes a channel matrix compressed using a first processing sub-model and identifiers of segmentation points corresponding to the terminal device. The first processing sub-model is determined based on a processing model. The identifiers of the segmentation points indicate that the segmentation point in the processing model and the portion before the segmentation point constitute the first processing sub-model, while the portion after the segmentation point constitutes a second processing sub-model. The second processing sub-model is used to decompress the compressed channel matrix. The compressed channel matrix is ​​then decompressed based on the second processing sub-model to obtain the channel matrix. The second processing sub-model is determined based on the segmentation points.

[0199] In one possible implementation, the segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

[0200] In one possible implementation, the processing model includes at least one processing layer, the processing model includes at least one segmentation point, a segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point; the portion of the processing model after the segmentation point is a second processing sub-model, including: the processing layer after the segmentation point in the processing model is the second processing sub-model.

[0201] In one possible implementation, the at least one segmentation point is determined based on the dimensional information and computational complexity information of each processing layer in the processing model; the dimensional information of a processing layer is used to indicate the dimension of the data output by the processing layer, and the computational complexity information of a processing layer is used to indicate the computational complexity required to run the processing sub-model corresponding to the processing layer; the processing sub-model corresponding to a processing layer is the processing layer and the part before the processing layer in the processing model.

[0202] In one possible implementation, the maximum number of the at least one segmentation point is determined by the number of bits occupied by the identifier of the segmentation point.

[0203] In one possible implementation, the chip module 1104 is further configured to: send the processing model and segmentation point information to the terminal device, wherein the segmentation point information includes the identifier of the at least one segmentation point.

[0204] In one possible implementation, the segmentation point information further includes the dimension information and computational cost information corresponding to the at least one segmentation point.

[0205] For various devices and products applied to or integrated into chip modules, each of its modules can be implemented using hardware methods such as circuits. Different modules can be located in the same component (e.g., chip, circuit module, etc.) or different components of the chip module. Alternatively, at least some modules can be implemented using software programs that run on the processor integrated inside the chip module, while the remaining (if any) modules can be implemented using hardware methods such as circuits.

[0206] This application also provides a computer-readable storage medium storing instructions that, when executed on a processor, enable the implementation of the method flow described in the above method embodiments.

[0207] This application also provides a computer program or computer program product, including code or instructions, which, when run on a computer, cause the computer to perform the methods described in the above method embodiments.

[0208] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some operations can be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0209] The descriptions of the various embodiments provided in this application can be referenced mutually. Each embodiment has its own emphasis, and parts not described in detail in a certain embodiment can be referred to the relevant descriptions of other embodiments. For the sake of convenience and brevity, for example, the functions and operations of the various devices and equipment provided in the embodiments of this application can be referred to the relevant descriptions of the method embodiments of this application. The method embodiments and the device embodiments can also be referenced, combined or cited from each other.

[0210] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. An information feedback method, characterized in that, Applied to a terminal device, the method includes: The channel matrix is ​​obtained by performing downlink channel estimation; The channel matrix is ​​compressed based on the first processing sub-model to obtain the compressed channel matrix; wherein, the first processing sub-model is determined according to the processing model, the segmentation point corresponding to the terminal device in the processing model and the part before the segmentation point are the first processing sub-model, the part after the segmentation point in the processing model are the second processing sub-model, and the second processing sub-model is used to decompress the compressed channel matrix. Channel state information is sent to the network device. The channel state information includes the compressed channel matrix and the identifier of the segmentation point, and the identifier of the segmentation point is used to indicate the segmentation point.

2. The method according to claim 1, characterized in that, The segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

3. The method according to claim 1 or 2, characterized in that, The processing model includes at least one processing layer, and the processing model includes at least one segmentation point. A segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point. The segmentation point corresponding to the terminal device in the processing model and the portion before the segmentation point constitute the first processing sub-model, including: The segmentation point corresponding to the terminal device in the processing model and the processing layer before the segmentation point constitute the first processing sub-model.

4. The method according to claim 3, characterized in that, The segmentation point is determined based on segmentation point information and first information, wherein the segmentation point information includes the identifier of at least one segmentation point of the processing model; The first information includes one or more of the following: capability information of the terminal device, channel quality of the channel between the terminal device and the network device; the capability information includes one or more of the following: computing power, communication power, and power information.

5. The method according to claim 4, characterized in that, The segmentation point information also includes the dimension information and computational cost information of the at least one segmentation point; The dimension information of a split point is used to indicate the dimension of the data output by that split point. The computational cost information of a split point is used to indicate the computational cost required to run the processing sub-model corresponding to that split point. The processing sub-model corresponding to a split point is the split point and the part before that split point in the processing model.

6. The method according to claim 4 or 5, characterized in that, The method further includes: Receive the segmentation point information from the network device.

7. The method according to claim 5, characterized in that, The dimensional information and computational complexity information of the at least one segmentation point are determined based on the identifier of the at least one segmentation point and the processing model.

8. The method according to any one of claims 1-7, characterized in that, The method further includes: The processing model is received from the network device.

9. An information feedback method, characterized in that, The method includes: The system receives channel state information, which includes a channel matrix compressed using a first processing sub-model and identifiers of segmentation points corresponding to the terminal device. The first processing sub-model is determined based on a processing model. The identifiers of the segmentation points indicate the segmentation points. The segmentation points and the portion preceding them in the processing model constitute the first processing sub-model, while the portion following them constitutes the second processing sub-model. The second processing model is used to decompress the compressed channel matrix. The compressed channel matrix is ​​decompressed based on the second processing sub-model to obtain the channel matrix; the second processing sub-model is determined based on the segmentation point.

10. The method according to claim 9, characterized in that, The segmentation point indicated by the segmentation point identifier is used to determine the data size of the compressed channel matrix.

11. The method according to claim 9 or 10, characterized in that, The processing model includes at least one processing layer, and the processing model includes at least one segmentation point. A segmentation point is a processing layer in the processing model, and the segmentation point is contained within the at least one segmentation point. The portion following the segmentation point in the processing model constitutes the second processing sub-model, which includes: The processing layer following the segmentation point in the processing model is the second processing sub-model.

12. The method according to claim 11, characterized in that, The at least one segmentation point is determined based on the dimensional information and computational complexity information of each processing layer in the processing model. The dimension information of a processing layer is used to indicate the dimension of the data output by the processing layer. The computational complexity information of a processing layer is used to indicate the computational complexity required to run the processing sub-model corresponding to the processing layer. The processing sub-model corresponding to a processing layer is the processing layer and the part before the processing layer in the processing model.

13. The method according to claim 12, characterized in that, The maximum number of the at least one segmentation point is determined by the number of bits occupied by the identifier of the segmentation point.

14. The method according to any one of claims 11-13, characterized in that, The method further includes: The processing model and segmentation point information are sent to the terminal device, wherein the segmentation point information includes the identifier of at least one segmentation point.

15. The method according to claim 14, characterized in that, The segmentation point information also includes the dimension information and computational cost information corresponding to the at least one segmentation point.

16. A communication device, characterized in that, Includes units for implementing the method of any one of claims 1-15.

17. A communication device, characterized in that, The communication device includes a processor and a transceiver; The transceiver is used to receive or send signals; The processor is configured to perform the method as described in any one of claims 1-15.

18. The communication device according to claim 17, characterized in that, The communication device also includes a memory: The memory is used to store computer programs; The processor is specifically configured to invoke the computer program from the memory, causing the communication device to perform the method as described in any one of claims 1-15.

19. A chip, characterized in that, The chip is used to perform downlink channel estimation to obtain the channel matrix; The chip is further configured to compress the channel matrix based on a first processing sub-model to obtain a compressed channel matrix; wherein, the first processing sub-model is determined according to a processing model, the segmentation point corresponding to the terminal device in the processing model and the part before the segmentation point are the first processing sub-model, the part after the segmentation point in the processing model are the second processing sub-model, and the second processing sub-model is used to decompress the compressed channel matrix; The chip is also used to send channel state information to network devices. The channel state information includes the compressed channel matrix and the identifier of the segmentation point, and the identifier of the segmentation point is used to indicate the segmentation point.

20. A chip, characterized in that, The chip is used to receive channel state information, which includes a channel matrix compressed using a first processing sub-model and identifiers of segmentation points corresponding to the terminal device. The first processing sub-model is determined based on a processing model, the identifiers of the segmentation points indicate the segmentation points, the segmentation points and the portion before the segmentation points in the processing model constitute the first processing sub-model, and the portion after the segmentation points in the processing model constitutes the second processing sub-model. The second processing model is used to decompress the compressed channel matrix. The chip is further configured to decompress the compressed channel matrix based on the second processing sub-model to obtain the channel matrix; the second processing sub-model is determined based on the segmentation point.

21. A module device, characterized in that, The module device includes a communication module, a power module, a storage module, and a chip module, wherein: The power module is used to provide electrical energy to the module device; The storage module is used to store data and instructions; The communication module is used for internal communication within the module device, or for communication between the module device and external devices; The chip module is used for: The channel matrix is ​​obtained by performing downlink channel estimation; The channel matrix is ​​compressed based on the first processing sub-model to obtain a compressed channel matrix; wherein, the first processing sub-model is determined according to the processing model, the segmentation point corresponding to the terminal device in the processing model and the part before the segmentation point are the first processing sub-model, the part after the segmentation point in the processing model are the second processing sub-model, and the second processing sub-model is used to decompress the compressed channel matrix. Channel state information is sent to the network device. The channel state information includes the compressed channel matrix and the identifier of the segmentation point, and the identifier of the segmentation point is used to indicate the segmentation point.

22. A module device, characterized in that, The module device includes a communication module, a power module, a storage module, and a chip module, wherein: The power module is used to provide electrical energy to the module device; The storage module is used to store data and instructions; The communication module is used for internal communication within the module device, or for communication between the module device and external devices; The chip module is used for: The system receives channel state information, which includes a channel matrix compressed using a first processing sub-model and identifiers of segmentation points corresponding to the terminal device. The first processing sub-model is determined based on a processing model. The identifiers of the segmentation points indicate the segmentation points. The segmentation points and the portion preceding them in the processing model constitute the first processing sub-model, while the portion following them constitutes the second processing sub-model. The second processing model is used to decompress the compressed channel matrix. The compressed channel matrix is ​​decompressed based on the second processing sub-model to obtain the channel matrix; the second processing sub-model is determined based on the segmentation point.

23. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions that, when executed on a communication device, cause the communication device to perform the method as described in any one of claims 1-15.

Images