Information processing method and apparatus, communication device, and storage medium

Abstract

This application discloses an information processing method, apparatus, communication device, and storage medium, relating to the field of wireless communication. The method includes: acquiring n first channel information corresponding to n feedback periods, where the feedback period is the feedback period of the channel information, and n is a positive integer greater than 1; performing m concatenation operations of the n first channel information at different scales to obtain m concatenated channel information, wherein the different scales of concatenation indicate that the number of first channel information segments concatenated in the m concatenated channel information are all different, and m is a positive integer; and inputting the m concatenated channel information into a neural network model for processing to obtain second channel information; wherein the neural network model is one of an encoding model and a decoding model. Embodiments of this application can enhance the encoding or decoding performance of channel information.

Description

Technical Field

[0001] This application relates to the field of wireless communication, and in particular to an information processing method, apparatus, communication device, and storage medium. Background Technology

[0002] In New Radio (NR) systems, channel information feedback is based on a codebook-based feedback scheme.

[0003] In codebook-based feedback schemes, the terminal device selects the optimal feedback matrix from the codebook based on the estimated channel. However, the codebook itself is finite, meaning that the mapping process from the estimated channel to the feedback matrix in the codebook is lossy in terms of quantization. This reduces the accuracy of the feedback channel information and consequently lowers the performance of precoding.

[0004] To address the issue of low accuracy in codebook-based feedback schemes, a neural network-based feedback scheme is proposed. In this scheme, the channel information obtained after channel estimation is encoded and compressed at the transmitting end, and then decoded and recovered at the receiving end. Summary of the Invention

[0005] This application provides an information processing method, apparatus, communication device, and storage medium, which can enhance the encoding or decoding performance of channel information. The technical solution is as follows:

[0006] According to one aspect of this application, an information processing method is provided, the method comprising:

[0007] Obtain n first channel information corresponding to n feedback periods, where the feedback period is the feedback period of the channel information, and n is a positive integer greater than 1;

[0008] The n first channel informations are spliced ​​together m times at different scales to obtain m spliced ​​channel informations. The different scales of splicing are used to indicate that the number of first channel informations spliced ​​together in the m spliced ​​channel informations are different from each other, and m is a positive integer.

[0009] The m concatenated channel information is input into a neural network model for processing to obtain the second channel information;

[0010] The neural network model is one of the encoding model and the decoding model.

[0011] According to one aspect of this application, an information processing apparatus is provided, the apparatus comprising: an information acquisition module, an information splicing module, and an information processing module;

[0012] The information acquisition module is used to acquire n first channel information corresponding to n feedback cycles, wherein the feedback cycle is the feedback cycle of the channel information, and n is a positive integer greater than 1;

[0013] The information splicing module is used to splice the n first channel informations m times at different scales to obtain m spliced ​​channel informations. The splicing at different scales is used to indicate that the number of first channel informations spliced ​​in the m spliced ​​channel informations is different from each other, and m is a positive integer.

[0014] The information processing module is used to input the m spliced ​​channel information into a neural network model for processing to obtain the second channel information;

[0015] The neural network model is one of the encoding model and the decoding model.

[0016] According to one aspect of this application, a terminal device is provided, the terminal device comprising: a processor; wherein...

[0017] The processor is configured to acquire n first channel information corresponding to n feedback cycles, wherein the feedback cycle is the feedback period of the channel information, and n is a positive integer greater than 1;

[0018] The processor is configured to perform m concatenations of the n first channel information at different scales to obtain m concatenated channel information. The concatenation at different scales is used to indicate that the number of first channel information concatenated in the m concatenated channel information is different from each other, where m is a positive integer.

[0019] The processor is used to input the m spliced ​​channel information into a neural network model for processing to obtain the second channel information;

[0020] The neural network model is an encoding model.

[0021] According to one aspect of this application, a network device is provided, the network device comprising: a processor and a transceiver connected to the processor; wherein,

[0022] The transceiver is used to acquire n first channel information corresponding to n feedback periods, wherein the feedback period is the feedback period of the channel information, and n is a positive integer greater than 1;

[0023] The processor is configured to perform m concatenations of the n first channel information at different scales to obtain m concatenated channel information. The concatenation at different scales is used to indicate that the number of first channel information concatenated in the m concatenated channel information is different from each other, where m is a positive integer.

[0024] The processor is used to input the m spliced ​​channel information into a neural network model for processing to obtain the second channel information;

[0025] The neural network model is a decoding model.

[0026] According to one aspect of this application, a computer-readable storage medium is provided, wherein executable instructions are stored therein, the executable instructions being loaded and executed by a processor to implement the information processing method as described above.

[0027] According to one aspect of the embodiments of this application, a chip is provided, the chip including programmable logic circuits and / or program instructions, which, when the chip is run on a computer device, are used to implement the information processing method described above.

[0028] According to one aspect of this application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium, wherein a processor reads from the computer-readable storage medium and executes the computer instructions to implement the information processing method described above.

[0029] The technical solutions provided in this application have at least the following beneficial effects:

[0030] The n first channel information corresponding to n feedback cycles are spliced ​​at different scales to obtain m spliced ​​channel information. Then, the m spliced ​​channel information is processed by a neural network model to obtain the second channel information. The neural network model is one of the encoding model and the decoding model, thereby realizing the multi-scale utilization of the first channel information in different feedback cycles and enhancing the encoding or decoding performance of the channel information. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0032] Figure 1 This is a schematic diagram of a network architecture provided in an exemplary embodiment of this application;

[0033] Figure 2 This is a schematic diagram of a neural network provided in an exemplary embodiment of this application;

[0034] Figure 3 This is a schematic diagram of a convolutional neural network provided in an exemplary embodiment of this application;

[0035] Figure 4 This is a schematic diagram of a long short-term memory network provided in an exemplary embodiment of this application;

[0036] Figure 5 This is a schematic diagram of channel information feedback based on artificial intelligence provided in an exemplary embodiment of this application;

[0037] Figure 6 This is a schematic diagram of a channel information feedback system provided in an exemplary embodiment of this application;

[0038] Figure 7 This is a schematic diagram illustrating channel recovery using historical feedback information, provided in an exemplary embodiment of this application.

[0039] Figure 8 This is a flowchart of an exemplary embodiment of the information processing method provided in this application;

[0040] Figure 9 This is a flowchart of an exemplary embodiment of the information processing method provided in this application;

[0041] Figure 10 This is a schematic diagram illustrating the enhancement of coding performance at the transmitting end based on multi-scale information, provided by an exemplary embodiment of this application.

[0042] Figure 11 This is a schematic diagram illustrating the enhancement of decoding performance at the receiving end based on multi-scale information, provided by an exemplary embodiment of this application.

[0043] Figure 12 This is a schematic diagram illustrating the enhancement of decoding performance at the receiving end based on multi-scale information, provided by an exemplary embodiment of this application.

[0044] Figure 13 This is a schematic diagram illustrating the enhancement of decoding performance at the receiving end based on multi-scale information, provided by an exemplary embodiment of this application.

[0045] Figure 14 This is a structural block diagram of an information processing apparatus provided in an exemplary embodiment of this application;

[0046] Figure 15 This is a schematic diagram of the structure of a communication device provided in an exemplary embodiment of this application. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0048] The network architecture and business scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0049] Please refer to Figure 1 This illustration shows a schematic diagram of a network architecture 100 provided in one embodiment of this application. The network architecture 100 may include: a terminal device 10, an access network device 20, and a core network device 30.

[0050] Terminal device 10 can refer to UE (User Equipment), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, wireless communication device, user agent, or user equipment. Optionally, terminal device 10 can also be a cellular phone, cordless phone, SIP (Session Initiation Protocol) phone, WLL (Wireless Local Loop) station, PDA (Personal Digital Assistant), handheld device with wireless communication function, computing device or other processing device connected to a wireless modem, vehicle-mounted device, wearable device, terminal device in 5GS (5th Generation System), or terminal device in the future evolved PLMN (Public Land Mobile Network), etc., and this application embodiment is not limited in this respect. For ease of description, the devices mentioned above are collectively referred to as terminal devices. The number of terminal devices 10 is usually multiple, and one or more terminal devices 10 can be distributed within the cell managed by each access network device 20.

[0051] Access network device 20 is a device deployed in an access network to provide wireless communication functionality to terminal device 10. Access network device 20 may include various forms of macro base stations, micro base stations, relay stations, access points, etc. In systems employing different wireless access technologies, the name of the device with access network device functionality may differ; for example, in a 5G NR system, it is called gNodeB or gNB. As communication technologies evolve, the name "access network device" may change. For ease of description, in this embodiment, the aforementioned devices providing wireless communication functionality to terminal device 10 are collectively referred to as access network devices. Optionally, a communication relationship can be established between terminal device 10 and core network device 30 through access network device 20. For example, in an LTE (Long Term Evolution) system, access network device 20 may be one or more eNodeBs in an EUTRAN (Evolved Universal Terrestrial Radio Access Network) or EUTRAN; in a 5G NR system, access network device 20 may be one or more gNBs in a RAN (Radio Access Network). In this embodiment of the application, unless otherwise specified, the network device refers to access network device 20, such as a base station.

[0052] Core network equipment 30 is equipment deployed in the core network. Its main functions are to provide user connectivity, manage users, and bear services, serving as an interface to external networks. For example, core network equipment in a 5G NR system may include AMF (Access and Mobility Management Function) entities, UPF (User Plane Function) entities, and SMF (Session Management Function) entities.

[0053] In one example, access network device 20 and core network device 30 communicate with each other via some air interface technology, such as the NG interface in a 5G NR system. Access network device 20 and terminal device 10 communicate with each other via some air interface technology, such as the Uu interface.

[0054] The "5G NR system" in this application embodiment can also be referred to as a 5G system or an NR system, but those skilled in the art will understand its meaning. The technical solutions described in this application embodiment can be applied to LTE systems, 5G NR systems, subsequent evolution systems of 5G NR systems, and other communication systems such as NB-IoT (Narrow Band Internet of Things) systems. This application does not limit these applications.

[0055] Before introducing the technical solution of this application, some background technical knowledge involved in this application will be introduced and explained.

[0056] For 5G NR systems, the current Channel State Information (CSI) feedback design primarily utilizes a codebook-based approach to extract and feedback channel features. Specifically, after channel estimation at the transmitting end, the best-matching precoding matrix is ​​selected from a pre-defined precoding codebook based on the estimation results and a certain optimization criterion. The matrix's index information is then fed back to the receiving end via the air interface feedback link, enabling the receiving end to perform precoding.

[0057] In recent years, artificial intelligence research, represented by neural networks, has achieved remarkable results in many fields, and it will continue to have a significant impact on people's production and lives for a long time to come.

[0058] Please refer to Figure 2 This illustrates a schematic diagram of a neural network provided in one embodiment of this application. Figure 2 As shown, the basic structure of a simple neural network includes an input layer, hidden layers, and an output layer. The input layer receives data, the hidden layers process the data, and the final result is generated in the output layer. Figure 1 As shown, each node represents a processing unit, or can be considered as simulating a neuron. Multiple neurons form a layer of neural network, and the transmission and processing of information in multiple layers construct a complete neural network.

[0059] With the continuous development of neural network research, deep learning algorithms for neural networks have been proposed in recent years, and more hidden layers have been introduced. By training neural networks with multiple hidden layers layer by layer to learn features, the learning and processing capabilities of neural networks have been greatly improved, and they have been widely used in pattern recognition, signal processing, optimization and combination, anomaly detection and other fields.

[0060] Meanwhile, with the development of deep learning, convolutional neural networks have also been further studied. Please refer to [link / reference needed]. Figure 3This illustrates a schematic diagram of a convolutional neural network provided in one embodiment of this application. Figure 3 As shown, the basic structure of a convolutional neural network includes an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, and an output layer. The introduction of convolutional and pooling layers effectively controls the rapid increase in network parameters, limits the number of parameters, and taps into the characteristics of local structures, thereby improving the robustness of the algorithm.

[0061] Recurrent Neural Networks (RNNs) are a type of recurrent neural network that takes sequential data as input and recursively processes the data in the direction of the sequence, with all nodes (recurrent units) connected in a chain-like manner. As the most commonly used and traditional deep learning model in Natural Language Processing (NLP), RNNs process sequential data step by step, much like how humans understand text—reading a book word by word, sentence by sentence.

[0062] Long Short-Term Memory (LSTM) networks are a variant of RNNs, such as... Figure 4 As shown, its essence lies in the introduction of the concept of cell state. Unlike RNN, which only considers the most recent state, LSTM's cell state determines which states should be retained and which should be forgotten, thus solving the shortcomings of traditional RNN in long-term memory.

[0063] Given the tremendous success of Artificial Intelligence (AI) technology in areas such as computer vision and natural language processing, the communications field has begun to explore new technological approaches to address technical challenges that are limited by traditional methods, such as deep learning. The neural network architecture commonly used in deep learning is non-linear and data-driven, capable of extracting features from actual channel matrix data and reconstructing the compressed channel matrix information from the terminal side as accurately as possible at the base station. This not only ensures the reconstruction of channel information but also provides the possibility of reducing the feedback overhead of channel information at the terminal side.

[0064] like Figure 5 As shown, in the AI-based channel information feedback, the channel information is treated as an image to be compressed 501. A deep learning autoencoder 502 is used to compress the channel information to obtain a compressed channel image 503. The receiver uses a deep learning autodecoder 504 to reconstruct the compressed channel image 503 to obtain the restored channel information 505, which can preserve the channel information to a greater extent.

[0065] A typical channel information feedback system is as follows: Figure 6 As shown, the entire feedback system is divided into an encoder and a decoder, deployed at the transmitting and receiving ends, respectively. After obtaining channel information through channel estimation, the transmitting end compresses and encodes the channel information matrix using the encoder's neural network, and then feeds the compressed bitstream back to the receiving end via the air interface feedback link. The receiving end uses the decoder to reconstruct the channel information based on the fed-back bitstream to obtain complete channel information. Figure 6 The structure shown uses several fully connected layers for encoding in the encoder and a residual network structure for decoding in the decoder. While keeping the encoding / decoding framework unchanged, the network model structures within the encoder and decoder can be flexibly designed.

[0066] Currently, the channel information feedback in the 5G NR standard uses a codebook-based feedback scheme. However, this scheme only selects the optimal feedback matrix from the codebook based on the estimated channel. The codebook itself has limitations; that is, the mapping process from the estimated channel to the channel in the codebook is lossy in terms of quantization. This reduces the accuracy of the feedback channel information, thereby lowering the performance of precoding.

[0067] To address the accuracy issues inherent in codebook-based feedback schemes, a neural network-based feedback scheme is discussed. This neural network-based channel information feedback scheme directly encodes and compresses the channel information obtained after channel estimation, thus alleviating the accuracy problems of codebook-based schemes.

[0068] The channel information fed back in different feedback periods has a certain historical correlation, which can be used to enhance the channel recovery performance of the current feedback period. That is, the channel information fed back in different feedback periods is used to form historical feedback information, which is an image or sequence, and is used as the input of the decoder.

[0069] Reference Figure 7 This diagram illustrates a channel recovery process utilizing historical feedback information. The entire feedback system includes an encoder at the transmitting end and a decoder at the receiving end.

[0070] The transmitting end compresses and encodes the channel information into a bit stream using an encoder within different feedback periods. In this embodiment, the maximum historical tracing scale is set to n. That is, in n feedback periods, the transmitting end encodes the channel information {H_1, ..., H_n} into a bit stream {B_1, ..., B_n} using the encoder, and transmits it to the receiving end via the air interface in each feedback period. When the receiving end decodes the feedback channel in the nth feedback period, it simultaneously uses the feedback bit streams from the previous n-1 feedback periods and the nth feedback period as input to the decoder. The output of the decoder network is the recovered channel H'_n.

[0071] If according to such Figure 7 The example illustrates channel recovery using historical feedback information. The length of the input historical feedback information directly impacts the network's performance after training. However, the length of the input historical feedback information is often set empirically, making it difficult to determine whether the current input length is optimal. Reflecting on the channel itself, the optimal length of the historical feedback information differs depending on the channel environment (e.g., different terminal movement speeds) when using historical feedback information for performance enhancement.

[0072] To address the aforementioned issues, the technical solution of this application employs the aforementioned approach of utilizing historical correlation and using multiple first channel information corresponding to multiple feedback periods to enhance the performance of the current feedback period at both the transmitting and receiving ends. Furthermore, the transmitting end (or receiving end) utilizes the first channel information from different feedback periods at multiple scales, thereby enhancing the coding or decoding performance corresponding to the channel information.

[0073] The technical solution of this application will be described and illustrated below through several embodiments.

[0074] Figure 8 A flowchart illustrating an exemplary embodiment of the information processing method provided in this application is shown. This method can be applied to, for example... Figure 1 In the network architecture shown, the method may include the following steps (802-806):

[0075] Step 802: Obtain n first channel information corresponding to n feedback periods, where the feedback period is the feedback period of the channel information.

[0076] Where n is a positive integer greater than 1. That is, the communication device acquires multiple first channel information corresponding to multiple feedback cycles.

[0077] In mobile communication systems, terminal devices need to periodically feed back channel information to network devices according to a feedback cycle, or periodically determine the channel information for different feedback cycles through channel estimation. Let the current feedback cycle be the nth feedback cycle, and going back n-1 feedback cycles, there are a total of n feedback cycles, including: the 1st feedback cycle, the 2nd feedback cycle, ..., the nth feedback cycle.

[0078] The first channel information is information related to the channel information feedback process.

[0079] For example, for a terminal device acting as the transmitter of channel information, it needs to perform channel estimation by measuring a reference signal to determine the channel information for different feedback periods. The first channel information is obtained through channel estimation. The terminal device determines n channel information corresponding to n feedback periods.

[0080] For example, for a network device acting as the receiver of channel information, it needs to receive compressed channel information from the transmitter for channel recovery. The first channel information is the compressed bit stream corresponding to the channel information. The network device receives the compressed bit stream corresponding to n channel information values ​​for n feedback periods.

[0081] Step 804: Perform m different scale splicing operations on n first channel information to obtain m spliced ​​channel information. The different scale splicing is used to indicate that the number of first channel information spliced ​​in the m spliced ​​channel information is different from each other.

[0082] To utilize the historical correlation among n pieces of first channel information, the n pieces of first channel information can be concatenated to obtain concatenated channel information.

[0083] In this embodiment, the communication device performs m concatenations on n pieces of first channel information. In each concatenation process, the first channel information from the n pieces of first channel information is concatenated, and the number of first channel information pieces concatenated in each concatenation is different; that is, the m concatenations are concatenations at m different scales. In this embodiment, m can be a positive integer not less than 2.

[0084] For example, n is 6, resulting in 6 first channel information items, including information 1 to information 6; m is 3, resulting in 3 spliced ​​channel information items. Spliced ​​channel information 1 splices information 1 to information 6, resulting in 6 first channel information items; spliced ​​channel information 2 splices information 3 to information 6, resulting in 3 first channel information items; spliced ​​channel information 3 splices information 5 and information 6, resulting in 2 first channel information items. The number of first channel information items spliced ​​in the above 3 spliced ​​channel information items is different for each other, which can be understood as 3 splicing operations at different scales.

[0085] It is understandable that the spliced ​​channel information may also include one first channel information. For example, the spliced ​​channel information includes: the nth first channel information corresponding to the nth feedback period.

[0086] Step 806: Input the m spliced ​​channel information into the neural network model for processing to obtain the second channel information.

[0087] On the one hand, since the spliced ​​channel information is composed of first channel information with different feedback periods, the first channel information with different feedback periods has a certain historical correlation, which can be used to enhance performance. On the other hand, since the number of first channel information pieces spliced ​​by the m spliced ​​channel information pieces is different, and the lengths of the m spliced ​​channel information pieces input to the neural network model are different, the first channel information can be utilized at multiple scales.

[0088] It is understood that the embodiments of this application do not limit the model structure of the neural network. For example, the model structure of the neural network includes, but is not limited to: fully connected neural networks, convolutional neural networks, recurrent neural networks, and long short-term memory networks.

[0089] In this application, the neural network model is one of the encoding and decoding models. In the neural network-based feedback scheme, the encoding model at the transmitting end and the decoding model at the receiving end are two mutually matched models. In this embodiment, the encoding model refers to the model used to encode channel information to generate a compressed bitstream; the decoding model refers to the model used to decode the received compressed bitstream to reconstruct the channel information. It can be understood that the encoding and decoding models can also be understood as: a channel state information encoding model and a channel state information decoding model; a channel encoding model and a channel decoding model; a modulation model and a demodulation model, etc., and this embodiment does not impose any limitations on this.

[0090] The second channel information is the output information after the neural network model processes the spliced ​​channel information at different scales.

[0091] For example, for a terminal device acting as the transmitter of channel information, it needs to compress and encode the channel information using an encoding model. The second channel information is then the compressed bit stream corresponding to the channel information. The terminal device processes m concatenated channel information bits into the encoding model to obtain the compressed bit stream corresponding to the channel information.

[0092] For example, for a network device acting as the receiver of channel information, it needs to decode the received compressed bitstream to reconstruct the channel information. The second channel information is then the reconstructed channel information. The network device obtains the reconstructed channel information by inputting m concatenated channel information segments into a decoding model for processing.

[0093] In summary, the technical solution of this application splices n first channel information corresponding to n feedback cycles at different scales to obtain m spliced ​​channel information. Then, a neural network model processes the m spliced ​​channel information to obtain second channel information. The neural network model is one of the encoding model and the decoding model, thereby realizing the multi-scale utilization of the first channel information at different feedback cycles and enhancing the encoding or decoding performance of the channel information.

[0094] In an illustrative embodiment, the communication device utilizes first channel information with different feedback periods at multiple scales based on granularity information.

[0095] Figure 9 A flowchart illustrating an exemplary embodiment of the information processing method provided in this application is shown. This method can be applied to, for example... Figure 1In the network architecture shown, the method may include the following steps (902-908):

[0096] Step 902: Obtain n first channel information corresponding to n feedback periods, where the feedback period is the feedback period of the channel information.

[0097] Where n is a positive integer greater than 1.

[0098] Optionally, n first channel information elements can form a sequence or feature map.

[0099] In one possible implementation, the communication device acquires n first channel information corresponding to n feedback cycles, and obtains n first channel information in sequence representation.

[0100] For example, n is 6, and there are 6 first channel information, including information 1 to information 6, forming the sequence {information 1, information 2, information 3, information 4, information 5, information 6}.

[0101] In another possible implementation, the communication device acquires n first channel information corresponding to n feedback cycles, processes the n first channel information through a first neural network layer, and obtains n first channel information represented by a feature map.

[0102] The first neural network layer is a neural network structure that supports representing the first channel information using feature maps. Optionally, the first neural network layer increases the dimensionality of the first channel information and transforms it into the dimension of a channel information matrix. Optionally, the first neural network layer includes fully connected layers.

[0103] For example, n is 6, and there are 6 first channel information, including information 1 to information 6. After being processed by the neural network layer, feature maps {information 1', information 2', information 3', information 4', information 5', and information 6'} are generated.

[0104] Step 904: Obtain granularity information, which is used to indicate granularity s.

[0105] Where s is a positive integer.

[0106] Optionally, s is a pre-set fixed value.

[0107] Optionally, s is a value that is adjusted according to different channel conditions. For example, if the current channel conditions require using as much spliced ​​channel information as possible, then the granularity s corresponds to a smaller value; if the current channel conditions do not require using as much spliced ​​channel information as possible, then the granularity s corresponds to a larger value.

[0108] Step 906: Based on granularity information, the n first channel informations are spliced ​​together m times at different scales to obtain m spliced ​​channel information.

[0109] Wherein, the difference in the number of first channel information segments spliced ​​in any two spliced ​​channel information segments is an integer multiple of s.

[0110] Since the granularity information indicates the granularity s, the communication device will splice the next channel information with s fewer first channel information than the previous spliced ​​channel information according to the granularity s.

[0111] Optionally, the feedback period corresponding to the first channel information spliced ​​in the spliced ​​channel information is continuous in the time dimension. That is, the communication device splices x first channel information corresponding to x consecutive feedback periods to obtain a spliced ​​channel information, where x is a positive integer.

[0112] Optionally, each of the m spliced ​​channel information pieces includes the nth first channel information corresponding to the nth feedback period. It can be understood that the nth feedback period is the current feedback period, and including the nth first channel information corresponding to the nth feedback period in the spliced ​​channel information allows the spliced ​​channel information to better reflect the current channel state.

[0113] Optionally, in the first channel information spliced ​​by the previous splicing channel information, the first s first channel information in the time dimension are removed, and the remaining first channel information is spliced ​​to obtain the next splicing channel information.

[0114] In one possible implementation, step 906 is implemented as follows: concatenate the (a-1)*s+1th to the nth first channel information from the nth first channel information to obtain the ath concatenated channel information, where a is a positive integer starting from 1 and increasing by one, and (a-1)*s+1 is less than n.

[0115] For example, if n is 10 and there are 10 first channel information, including information 1 to information 10, and s is 3, then the spliced ​​channel information includes: {information 1, information 2, ..., information 10}, {information 4, information 5, ..., information 10}, {information 7, information 8, information 9, information 10}, and {information 10}.

[0116] Step 908: Input the m spliced ​​channel information into the neural network model for processing to obtain the second channel information.

[0117] In one possible implementation, the neural network model optimally combines the m concatenated channel information. That is, step 908 is replaced by: inputting the m concatenated channel information into m second neural network layers for processing to obtain m channel features; weighting and concatenating the m channel features to obtain concatenated channel features; and inputting the concatenated channel features into a third neural network layer to obtain second channel information.

[0118] In another possible implementation, the neural network model adaptively selects the optimal use of the m concatenated channel information. That is, step 908 is replaced by: inputting the m concatenated channel information into m second neural network layers for processing to obtain m channel features; selecting the target channel feature from the m channel features, and inputting the target channel feature into the fourth neural network layer to obtain the second channel information.

[0119] This application does not limit the specific implementation of the second, third, and fourth neural network layers.

[0120] In summary, the technical solution of this application splices n first channel information corresponding to n feedback cycles at different scales to obtain m spliced ​​channel information. Then, a neural network model processes the m spliced ​​channel information to obtain second channel information. The neural network model is one of the encoding model and the decoding model, thereby realizing the multi-scale utilization of the first channel information at different feedback cycles and enhancing the encoding or decoding performance of the channel information.

[0121] The technical solution of this application splices n first channel information at different scales based on granularity information. Since the granularity information indicates that the difference between the number of first channel information spliced ​​in any two spliced ​​channel information is an integer multiple of the granularity s, it can ensure that a reasonable number of spliced ​​channel information is obtained.

[0122] The technical solution of this application supports adaptive optimal selection or optimal joint utilization of spliced ​​channel information of different scales.

[0123] Based on the above embodiments, both the transmitting and receiving ends can employ the multi-scale splicing channel information scheme described above to enhance performance. Specifically, this includes the following two scenarios:

[0124] • The response to the neural network model is an encoding model. The first channel information includes the channel information obtained through channel estimation, and the second channel information includes the compressed bit stream corresponding to the channel information.

[0125] That is, at the transmitting end of the terminal device, a coding model is set up to enhance the coding performance of the compression process of the current feedback period by using channel information of different scales.

[0126] • The response to the neural network model is a decoding model. The first channel information includes the compressed bit stream corresponding to the channel information, and the second channel information includes the restored channel information.

[0127] In other words, a decoding model is set up at the receiving end of the network device to enhance the decoding performance of the decompression process in the current feedback cycle by using compressed bit streams of different scales.

[0128] The technical solution of this application will be illustrated by the following examples.

[0129] Reference Figure 10 This illustration shows a schematic diagram of encoding performance enhancement based on multi-scale information at the transmitting end, provided by an exemplary embodiment of this application.

[0130] In this embodiment, the channel information for n feedback cycles is denoted as H_1 to H_n, and the compressed bit stream B_n corresponding to the current nth feedback cycle is output. This embodiment corresponds to a joint enhancement mechanism based on LSTM. This embodiment treats combinations of channel information at different scales as sequences of different lengths as input to the encoder.

[0131] First, the channel information {H_1, ..., H_n} for n feedback cycles is flattened to compress its dimensions, resulting in {H'_1, ..., H'_n}. The granularity information s is set to 1, and the maximum historical scale is n, resulting in n different scales. The encoder requires n LSTM structures.

[0132] Treat {H'_1, ..., H'_n} and {H'_2, ..., H'_n} to {H'_n} as sequences, and use them as inputs to the 1st to nth LSTM structures respectively. At the same time, each LSTM structure only outputs the network output of the last loop of the LSTM.

[0133] Furthermore, the outputs {R_1, ..., R_n} of each LSTM structure are concatenated along the channel dimension and then amplified in dimension through a fully connected layer to convert the information into a compressed bitstream B_n corresponding to the current feedback cycle.

[0134] In this embodiment, the neural network model design at the transmitting end utilizes multi-scale optimal joint utilization of channel information for different feedback periods to enhance coding performance.

[0135] Reference Figure 11 This illustration shows a schematic diagram of a decoding performance enhancement at the receiving end based on multi-scale information, provided by an exemplary embodiment of this application.

[0136] In this embodiment, the compressed bitstreams of n feedback cycles are denoted as B_1 to B_n, and the restored channel information H'_n corresponding to the current nth feedback cycle is output. This embodiment corresponds to a joint enhancement mechanism based on a convolutional neural network. This embodiment treats combinations of channel information at different scales as images of different lengths, which are then used as input to the decoder.

[0137] Figure 11 (a) in the diagram corresponds to an encoder structure. First, the input channel information is converted into a one-dimensional vector H_n and input to the model. The model uses a fully connected neural network, including M fully connected layers. The last fully connected layer converts the information into a compressed bitstream B_n, where M is a positive integer. Activation layers, normalization layers, quantization layers, and other network layers can also be added between the fully connected layers.

[0138] Figure 11 (b) in the diagram corresponds to one decoder structure. A compressed bitstream {B_1, ..., B_n} with n feedback cycles is used as the decoder input. Each compressed bitstream first passes through a fully connected layer to enlarge its dimensions and convert them into the dimensions of the channel information matrix, generating a feature map {B'_1, ..., B'_n}. The granularity information s is set to 1, and the maximum historical scale is n, resulting in a total of n different scales. The decoder requires n residual block structures.

[0139] {B'_1, ..., B'_n}, {B'_2, ..., B'_n} to {B'_n} are concatenated along the channel dimension and used as inputs to the first to the nth residual block structures, respectively. Further, the outputs {R_1, ..., R_n} of each residual block are concatenated along the channel dimension and merged using a 1x1 convolutional layer. Finally, the residual blocks are used for reconstruction to obtain the restored channel information H'_n.

[0140] In this embodiment, the neural network model design at the receiving end utilizes multi-scale optimal joint utilization of compressed bitstreams of channel information with different feedback periods to enhance decoding performance.

[0141] Reference Figure 12 This illustration shows a schematic diagram of a decoding performance enhancement at the receiving end based on multi-scale information, provided by an exemplary embodiment of this application.

[0142] In this embodiment, the compressed bitstreams for n feedback cycles are denoted as B_1 to B_n, and the restored channel information H'_n corresponding to the current nth feedback cycle is output. This embodiment corresponds to a joint enhancement mechanism based on a recurrent neural network. This embodiment treats combinations of compressed bitstreams of different scales as sequences of different lengths as input to the decoder.

[0143] Figure 12 (a) in the diagram corresponds to an encoder structure. The feature extractor uses a convolutional neural network, consisting of M convolutional neural network layers, with the last fully connected layer converting the information into a compressed bitstream B_n, where M is a positive integer. Activation layers, normalization layers, quantization layers, and other network layers can also be added between the various neural network layers.

[0144] Figure 12 (b) in the diagram corresponds to one type of decoder structure. A compressed bitstream {B_1, ..., B_n} with n feedback cycles is used as the decoder input. The granularity information s is set to 1, and the maximum historical scale is n, resulting in a total of n different scales. The decoder requires n RNN structures.

[0145] Treat {B_1, ..., B_n} and {B_2, ..., B_n} to {B_n} as sequences, and use them as inputs to the first to the nth RNN structures respectively. At the same time, each RNN structure only outputs the network output of the last loop of the RNN.

[0146] Furthermore, the outputs {R_1, ..., R_n} of each RNN structure are concatenated according to the channel dimension, and then merged using a 1x1 convolutional layer. The dimension is then amplified and converted into the dimension of the channel information matrix through a fully connected layer. Finally, the channel information H'_n is reconstructed and restored through residual blocks.

[0147] In this embodiment, the neural network model design at the receiving end utilizes multi-scale optimal joint utilization of compressed bitstreams of channel information with different feedback periods to enhance decoding performance.

[0148] Reference Figure 13 This illustration shows a schematic diagram of a decoding performance enhancement at the receiving end based on multi-scale information, provided by an exemplary embodiment of this application.

[0149] In this embodiment, the compressed bitstreams for n feedback cycles are denoted as B_1 to B_n, and the restored channel information H'_n corresponding to the current nth feedback cycle is output. This embodiment corresponds to an adaptive selection mechanism based on a long short-term memory network. This embodiment treats combinations of compressed bitstreams of different scales as sequences of different lengths as input to the decoder.

[0150] Figure 13 (a) in the diagram corresponds to an encoder structure. It employs a convolutional neural network and the Inception architecture, which uses different convolutional kernel sizes to extract features from the channel information H_n, concatenates the feature maps along the channel dimension, merges the feature maps using 1x1 convolutional layers, and finally converts the information into an output compressed bitstream B_n through a fully connected layer. Simultaneously, activation layers, normalization layers, quantization layers, and other network layers can be added between the various neural network layers.

[0151] Figure 13(b) in the diagram corresponds to one type of decoder structure. A compressed bitstream {B_1, ..., B_n} with n feedback cycles is used as the decoder input. The granularity information s is set to 1, and the maximum historical scale is n, resulting in a total of n different scales. The decoder requires n LSTM structures.

[0152] Treat {B_1, ..., B_n} and {B_2, ..., B_n} to {B_n} as sequences, and use them as inputs to the 1st to nth LSTM structures respectively. At the same time, each LSTM structure only outputs the network output of the last loop of the LSTM.

[0153] Furthermore, the outputs {R_1, ..., R_n} of each LSTM structure are concatenated into a tensor P along the channel dimension. This tensor P is then merged using a 1x1 convolutional layer. Simultaneously, two fully connected layers output a one-hot selection vector, which is then multiplied by the tensor P along the channel dimension, completing the selection operation from multiple branches to a single branch. Next, a fully connected layer enlarges the dimension and transforms it into the dimension of the channel information matrix. Finally, residual blocks are used for reconstruction to obtain the restored channel information H'_n.

[0154] In this embodiment, the neural network model design at the receiving end performs multi-scale adaptive selection and utilization of compressed bit streams of channel information with different feedback periods to enhance decoding performance.

[0155] It is understood that the neural network model structure shown in the above examples does not constitute a limitation on the technical solution of this application. Different data characteristics or channel characteristics will have different impacts on the selection of the above models. That is, the model selection needs to match the current data characteristics or channel characteristics. Neural network models can be adjusted accordingly based on different channel data, for example: modifying the encoder's feature extractor to other feature extraction networks, replacing LSTM with other forms of recurrent neural network modules, etc.

[0156] It should be noted that the above method embodiments can be implemented individually or in combination, and this application does not impose any restrictions on this.

[0157] In the above embodiments, the steps performed by the terminal device can be implemented independently as an information processing method on the terminal device side, and the steps performed by the network device can be implemented independently as an information processing method on the network device side.

[0158] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0159] Figure 14The diagram shows a structural block diagram of an information processing apparatus provided in an exemplary embodiment of this application. The apparatus can be implemented as a communication device, or as part of a communication device. The apparatus includes: an information acquisition module 1401, an information splicing module 1402, and an information processing module 1403.

[0160] The information acquisition module 1401 is used to acquire n first channel information corresponding to n feedback cycles, wherein the feedback cycle is the feedback cycle of the channel information, and n is a positive integer greater than 1;

[0161] The information splicing module 1402 is used to splice the n first channel informations at different scales m times to obtain m spliced ​​channel informations. The splicing at different scales is used to indicate that the number of first channel informations spliced ​​in the m spliced ​​channel informations is different from each other, and m is a positive integer.

[0162] The information processing module 1403 is used to input the m spliced ​​channel information into a neural network model for processing to obtain the second channel information;

[0163] The neural network model is one of the encoding model and the decoding model.

[0164] In an optional embodiment, the information splicing module 1402 includes: a granular information acquisition submodule and an information splicing submodule;

[0165] The granularity information acquisition submodule is used to acquire granularity information, which is used to indicate granularity s, where s is a positive integer;

[0166] The information splicing submodule is used to splice the n first channel informations at different scales m times based on the granularity information to obtain the m spliced ​​channel information, wherein the difference in the number of first channel informations spliced ​​in any two spliced ​​channel informations is an integer multiple of s.

[0167] In an optional embodiment, each of the m spliced ​​channel information includes the nth first channel information corresponding to the nth feedback period.

[0168] In an optional embodiment, the information splicing submodule is used to splice the (a-1)*s+1th to the nth first channel information from the n first channel information to obtain the ath spliced ​​channel information, where a is a positive integer starting from 1 and increasing by one, and (a-1)*s+1 is less than n.

[0169] In an optional embodiment, the information acquisition module 1401 is used to acquire n first channel information corresponding to n feedback cycles, and obtain the n first channel information in sequence representation;

[0170] or,

[0171] The information acquisition module 1401 is used to acquire n first channel information corresponding to n feedback cycles, and process the n first channel information through a first neural network layer to obtain the n first channel information represented by a feature map.

[0172] In an optional embodiment, the information processing module 1403 is configured to: input the m concatenated channel information into m second neural network layers for processing to obtain m channel features; perform weighted concatenation of the m channel features to obtain concatenated channel features; input the concatenated channel features into a third neural network layer to obtain the second channel information; or, select a target channel feature from the m channel features and input the target channel feature into a fourth neural network layer to obtain the second channel information.

[0173] In an optional embodiment, in response to the neural network model being an encoding model, the first channel information includes channel information obtained through channel estimation, and the second channel information includes a compressed bitstream corresponding to the channel information.

[0174] In an optional embodiment, in response to the neural network model being a decoding model, the first channel information includes a compressed bitstream corresponding to the channel information, and the second channel information includes the restored channel information.

[0175] It should be noted that the device provided in the above embodiments is only illustrated by the division of the above functional modules when implementing its functions. In actual applications, the above functions can be assigned to different functional modules according to actual needs, that is, the content structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0176] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0177] Figure 15 The diagram shows a schematic representation of the structure of a communication device (terminal device or network device) provided in an exemplary embodiment of this application. The communication device includes: a processor 1501, a receiver 1502, a transmitter 1503, a memory 1504, and a bus 1505.

[0178] The processor 1501 includes one or more processing cores, and the processor 1501 executes various functional applications and information processing by running software programs and modules.

[0179] The receiver 1502 and the transmitter 1503 can be implemented as a communication component, which can be a communication chip.

[0180] The memory 1504 is connected to the processor 1501 via the bus 1505.

[0181] The memory 1504 can be used to store at least one instruction, and the processor 1501 can execute the at least one instruction to implement the various steps in the above method embodiments.

[0182] Furthermore, the memory 1504 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, including but not limited to: magnetic disks or optical disks, electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), read-only memory (ROM), magnetic storage, flash memory, and programmable read-only memory (PROM).

[0183] When the communication device is implemented as a terminal device, the processor and transceiver involved in the embodiments of this application can perform the above-mentioned functions. Figures 8 to 10 The steps performed by the terminal device in any of the methods shown will not be described in detail here.

[0184] In one possible implementation, when the communication device implements a terminal device,

[0185] The processor is configured to acquire n first channel information corresponding to n feedback cycles, wherein the feedback cycle is the feedback period of the channel information, and n is a positive integer greater than 1;

[0186] The processor is configured to perform m concatenations of the n first channel information at different scales to obtain m concatenated channel information. The concatenation at different scales is used to indicate that the number of first channel information concatenated in the m concatenated channel information is different from each other, where m is a positive integer.

[0187] The processor is used to input the m spliced ​​channel information into a neural network model for processing to obtain the second channel information;

[0188] The neural network model is an encoding model.

[0189] When the communication device is implemented as a network device, the processor and transceiver involved in the embodiments of this application can perform the above-mentioned functions. Figures 8 to 9 , Figures 11 to 13 The steps performed by the network device in any of the methods shown will not be described in detail here.

[0190] In one possible implementation, when the communication device is implemented as a network device,

[0191] The transceiver is used to acquire n first channel information corresponding to n feedback periods, wherein the feedback period is the feedback period of the channel information, and n is a positive integer greater than 1;

[0192] The processor is configured to perform m concatenations of the n first channel information at different scales to obtain m concatenated channel information. The concatenation at different scales is used to indicate that the number of first channel information concatenated in the m concatenated channel information is different from each other, where m is a positive integer.

[0193] The processor is used to input the m spliced ​​channel information into a neural network model for processing to obtain the second channel information;

[0194] The neural network model is a decoding model.

[0195] This application also provides a computer-readable storage medium storing a computer program, which is executed by the processor of a terminal device to implement the above-described information processing method on the terminal device side.

[0196] This application also provides a computer-readable storage medium storing a computer program that is executed by a processor of a network device to implement the above-described information processing method on the network device side.

[0197] Optionally, the computer-readable storage medium may include: ROM (Read-Only Memory), RAM (Random-Access Memory), SSD (Solid State Drives), or optical disc, etc. The random access memory may include ReRAM (Resistance Random Access Memory) and DRAM (Dynamic Random Access Memory).

[0198] This application also provides a chip, which includes programmable logic circuits and / or program instructions. When the chip is run on a terminal device, it is used to implement the above-mentioned information processing method on the terminal device side.

[0199] This application also provides a chip, which includes programmable logic circuits and / or program instructions. When the chip is run on a network device, it is used to implement the above-described information processing method on the network device side.

[0200] This application also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of a terminal device reads and executes the computer instructions from the computer-readable storage medium to implement the aforementioned information processing method on the terminal device side.

[0201] This application also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of a network device reads and executes the computer instructions from the computer-readable storage medium to implement the above-described information processing method on the network device side.

[0202] It should be understood that the term "instruction" mentioned in the embodiments of this application can be a direct instruction, an indirect instruction, or an indication of a relationship. For example, A instructing B can mean that A directly instructs B, such as B being able to obtain information through A; it can also mean that A indirectly instructs B, such as A instructing C, so B can obtain information through C; or it can mean that there is a relationship between A and B.

[0203] In the description of the embodiments of this application, the term "correspondence" may indicate that there is a direct or indirect correspondence between two things, or that there is an association between two things, or that there is a relationship of instruction and being instructed, configuration and being configured, etc.

[0204] In this article, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0205] Furthermore, the step numbers described herein are merely illustrative of one possible execution order between steps. In some other embodiments, the steps may not be executed in the order of their numbers, such as two steps with different numbers being executed simultaneously, or two steps with different numbers being executed in the reverse order of the illustration. This application does not limit this.

[0206] Those skilled in the art will recognize that the functions described in the embodiments of this application in one or more of the above examples can be implemented using hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transfer of a computer program from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0207] The above description is merely an exemplary embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An information processing method, characterized in that, The method includes: Obtain n first channel information corresponding to n feedback periods, where the feedback period is the feedback period of the channel information, and n is a positive integer greater than 1; The n first channel informations are spliced ​​together m times at different scales to obtain m spliced ​​channel informations. The m different scale splicing means that the number of first channel informations spliced ​​together in the m spliced ​​channel informations are all different, and m is a positive integer. The m concatenated channel information is input into a neural network model for processing to obtain the second channel information; The neural network model is one of the encoding model and the decoding model. in: In response to the neural network model being an encoding model, the first channel information includes channel information obtained through channel estimation, and the second channel information includes the compressed bitstream corresponding to the channel information; In response to the neural network model being a decoding model, the first channel information includes the compressed bitstream corresponding to the channel information, and the second channel information includes the restored channel information.

2. The method according to claim 1, characterized in that, The process of concatenating the n first channel information pieces at different scales m times to obtain m concatenated channel information pieces includes: Obtain granularity information, which is used to indicate granularity s, where s is a positive integer; Based on the granularity information, the n first channel informations are spliced ​​together m times at different scales to obtain the m spliced ​​channel informations. The difference in the number of first channel informations spliced ​​in any two spliced ​​channel informations is an integer multiple of s.

3. The method according to claim 2, characterized in that, The m spliced ​​channel information each includes the nth first channel information corresponding to the nth feedback period.

4. The method according to claim 3, characterized in that, Based on the granularity information, the n first channel information pieces are spliced ​​together m times at different scales to obtain the m spliced ​​channel information pieces, including: The (a-1)th of the n first channel information The first channel information from s+1 to the nth first channel information is concatenated to obtain the ath concatenated channel information, where a is a positive integer starting from 1 and incrementing by one, and (a-1) s+1 is less than n.

5. The method according to any one of claims 1 to 4, characterized in that, The acquisition of n first channel information corresponding to n feedback cycles includes: Obtain n first channel information corresponding to n feedback cycles to obtain the n first channel information in sequence representation; or, Obtain n first channel information corresponding to n feedback cycles, process the n first channel information through a first neural network layer to obtain the n first channel information represented by a feature map.

6. The method according to any one of claims 1 to 4, characterized in that, The step of inputting the m concatenated channel information into a neural network model for processing to obtain the second channel information includes: The m spliced ​​channel information is input into m second neural network layers for processing to obtain m channel features; The m channel features are weighted and concatenated to obtain concatenated channel features. The concatenated channel features are then input into the third neural network layer to obtain the second channel information. or, Select the target channel feature from the m channel features, and input the target channel feature into the fourth neural network layer to obtain the second channel information.

7. An information processing device, characterized in that, The device includes: an information acquisition module, an information splicing module, and an information processing module; The information acquisition module is used to acquire n first channel information corresponding to n feedback cycles, wherein the feedback cycle is the feedback cycle of the channel information, and n is a positive integer greater than 1; The information splicing module is used to splice the n first channel informations at m different scales to obtain m spliced ​​channel informations. The m different scale splicing means that the number of first channel informations spliced ​​in the m spliced ​​channel informations are all different, and m is a positive integer. The information processing module is used to input the m spliced ​​channel information into a neural network model for processing to obtain the second channel information; The neural network model is one of the encoding model and the decoding model. in: In response to the neural network model being an encoding model, the first channel information includes channel information obtained through channel estimation, and the second channel information includes the compressed bitstream corresponding to the channel information; In response to the neural network model being a decoding model, the first channel information includes the compressed bitstream corresponding to the channel information, and the second channel information includes the restored channel information.

8. The apparatus according to claim 7, characterized in that, The information splicing module includes: a granular information acquisition submodule and an information splicing submodule; The granularity information acquisition submodule is used to acquire granularity information, which is used to indicate granularity s, where s is a positive integer; The information splicing submodule is used to splice the n first channel informations at different scales m times based on the granularity information to obtain the m spliced ​​channel information, wherein the difference in the number of first channel informations spliced ​​in any two spliced ​​channel informations is an integer multiple of s.

9. The apparatus according to claim 8, characterized in that, The m spliced ​​channel information each includes the nth first channel information corresponding to the nth feedback period.

10. The apparatus according to claim 9, characterized in that, The information splicing submodule is used to splice the (a-1)th of the n first channel information. The first channel information from s+1 to the nth first channel information is concatenated to obtain the ath concatenated channel information, where a is a positive integer starting from 1 and incrementing by one, and (a-1) s+1 is less than n.

11. The apparatus according to any one of claims 7 to 10, characterized in that, The information acquisition module is used to acquire n first channel information corresponding to n feedback cycles, and obtain the n first channel information in sequence representation; or, The information acquisition module is used to acquire n first channel information corresponding to n feedback cycles, and process the n first channel information through a first neural network layer to obtain the n first channel information represented by a feature map.

12. The apparatus according to any one of claims 7 to 10, characterized in that, The information processing module is used for: The m spliced ​​channel information is input into m second neural network layers for processing to obtain m channel features; The m channel features are weighted and concatenated to obtain concatenated channel features. The concatenated channel features are then input into the third neural network layer to obtain the second channel information. or, Select the target channel feature from the m channel features, and input the target channel feature into the fourth neural network layer to obtain the second channel information.

13. A terminal device, characterized in that, The terminal device includes: a processor; wherein... The processor is configured to acquire n first channel information corresponding to n feedback cycles, wherein the feedback cycle is the feedback period of the channel information, and n is a positive integer greater than 1; The processor is configured to perform m different scale splicing operations on the n first channel information to obtain m spliced ​​channel information, wherein the m different scale splicing operations refer to the fact that the number of first channel information spliced ​​in the m spliced ​​channel information is different from each other, and m is a positive integer; The processor is used to input the m spliced ​​channel information into a neural network model for processing to obtain the second channel information; The neural network model is an encoding model, the first channel information includes channel information obtained through channel estimation, and the second channel information includes the compressed bit stream corresponding to the channel information.

14. A network device, characterized in that, The network device includes: a processor and a transceiver connected to the processor; wherein, The transceiver is used to acquire n first channel information corresponding to n feedback periods, wherein the feedback period is the feedback period of the channel information, and n is a positive integer greater than 1; The processor is configured to perform m different scale splicing operations on the n first channel information to obtain m spliced ​​channel information, wherein the m different scale splicing operations refer to the fact that the number of first channel information spliced ​​in the m spliced ​​channel information is different from each other, and m is a positive integer; The processor is used to input the m spliced ​​channel information into a neural network model for processing to obtain the second channel information; The neural network model is a decoding model, the first channel information includes the compressed bit stream corresponding to the channel information, and the second channel information includes the restored channel information.

15. A computer-readable storage medium, characterized in that, The readable storage medium stores executable instructions, which are loaded and executed by a processor to implement the information processing method as described in any one of claims 1 to 6.

16. A chip, characterized in that, The chip includes programmable logic circuits and / or program instructions, which, when the chip is running, are used to implement the information processing method as described in any one of claims 1 to 6.

17. A computer program product or computer program, characterized in that, The computer program product or computer program includes computer instructions stored in a computer-readable storage medium, and a processor reads from and executes the computer instructions to implement the information processing method as described in any one of claims 1 to 6.

Images