Channel state information feedback method and device, electronic medium and equipment

Abstract

The embodiment of the invention discloses a channel state information feedback method and device, an electronic medium and equipment. The method comprises the following steps: acquiring an original channel state information matrix, and performing spatial dimension compression and time dimension compression on the original channel state information to obtain a specified channel state information matrix; performing coding processing on the specified channel state information matrix through a joint source channel coding technology to obtain a coded channel state information matrix; performing Gaussian noise superposition processing on the coded channel information state matrix according to a channel quality state to obtain a superposed channel state information matrix; and decoding the superposed channel state information matrix through the joint source channel coding technology to obtain target channel state information, and feeding back the target channel state information to a user side.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to a method, apparatus, electronic medium, and device for feedback of channel state information. Background Technology

[0002] Channel State Information (CSI) refers to the channel attributes of a communication link. It describes the signal attenuation factor on each transmission path. CSI enables the communication system to adapt to current channel conditions, ensuring high-reliability and high-speed communication in multi-antenna systems.

[0003] In related technologies, during the channel state information feedback process, spherical waves cause the channel path to spread in the angular domain. This leads to a single near-field path component spreading to multiple angles, thereby disrupting the sparsity of the channel in the angular domain. Summary of the Invention

[0004] This application / disclosure addresses some deficiencies mentioned in the background art by providing a method, apparatus, electronic medium, and device for feedback of channel state information.

[0005] In a first aspect, embodiments of this disclosure provide a method for feeding back channel state information, comprising: Obtain the original channel state information matrix, and compress the original channel state information in both spatial and temporal dimensions to obtain the specified channel state information matrix; The specified channel state information matrix is ​​encoded using joint source-channel coding techniques to obtain the encoded channel state information matrix. Based on the channel quality status, the encoded channel information state matrix is ​​subjected to Gaussian noise superposition processing to obtain the superimposed channel state information matrix. The superimposed channel state information matrix is ​​decoded using the joint source-channel coding technique to obtain the target channel state information, which is then fed back to the user terminal.

[0006] In one embodiment of the first aspect, the step of performing spatial and temporal compression on the original channel state information to obtain a specified channel state information matrix includes: The original channel state information matrix is ​​compressed in the spatial dimension to obtain an intermediate channel state information matrix; The intermediate channel state information matrix is ​​compressed in the time dimension to obtain the specified channel state information matrix.

[0007] In one embodiment of the first aspect, the step of compressing the original channel state information matrix in the spatial dimension to obtain an intermediate channel state information matrix includes: Based on formula Determine the intermediate channel state information matrix; Wherein, H is the original channel state information matrix. M is the number of antennas in the original channel, N is the number of subcarriers, and W... AD For angle-distance codebook, K AD These are the basis vectors bound by angle and distance.

[0008] In one embodiment of the first aspect, the step of compressing the intermediate channel state information matrix in the time dimension to obtain a specified channel state information matrix includes: Based on formula Determine the specified channel state information matrix; in, W is the intermediate channel state information matrix. TD The time-delay domain DFT transformation matrix, Nc represents the number of dominant delay paths.

[0009] In one embodiment of the first aspect, the step of encoding the specified channel state information matrix using joint source-channel coding techniques to obtain an encoded channel state information matrix includes: Based on formula Encoding is performed to obtain the encoded channel state information matrix; Among them, f en ( ) represents the compression function of the joint source-channel coding technique, θ en H represents the parameters of the neural network. s This specifies the channel state information matrix.

[0010] In one embodiment of the first aspect, the step of performing Gaussian noise superposition processing on the encoded channel state information matrix according to the channel quality state to obtain a superimposed channel state information matrix includes: Obtain the signal-to-noise ratio parameter; Based on the channel quality state and the signal-to-noise ratio parameter, determine the target Gaussian noise of the specified channel state information matrix; The target Gaussian noise is superimposed onto the encoded channel state information matrix to obtain the superimposed channel state information matrix.

[0011] In one embodiment of the first aspect, the step of decoding the superimposed channel state information matrix using the joint source-channel coding technique to obtain target channel state information includes: The superimposed channel state information matrix is ​​subjected to multi-scale feature enhancement processing to obtain the first superimposed matrix; The first superimposed matrix is ​​subjected to progressive convolution to obtain the second superimposed matrix; The target channel state information matrix is ​​obtained by processing the second superimposed matrix using an activation function. The target channel state information matrix is ​​restored to obtain the target channel state information.

[0012] In a second aspect, embodiments of this disclosure provide a channel state information feedback device, comprising: A compression unit is used to acquire the original channel state information matrix and perform spatial and temporal compression on the original channel state information to obtain a specified channel state information matrix. The encoding unit is used to encode the specified channel state information matrix using joint source-channel coding techniques to obtain the encoded channel state information matrix. The noise superposition unit is used to perform Gaussian noise superposition processing on the encoded channel information state matrix according to the channel quality state to obtain the superimposed channel state information matrix. The decoding unit is used to decode the superimposed channel state information matrix using the joint source channel coding technique to obtain the target channel state information, and to feed the target channel state information back to the user terminal.

[0013] In a third aspect, embodiments of this disclosure also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the first aspect or any possible implementation thereof.

[0014] Fourthly, embodiments of this disclosure also provide an electronic device, including: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and when the machine-readable instructions are executed by the processor, the steps of the first aspect above, or any possible implementation of the first aspect, are performed.

[0015] This disclosure has at least the following beneficial effects: By using spatial-temporal dual-dimensional compression, the spatial dimension can capture the near-field CSI coupling characteristics based on the angle-distance joint representation, avoiding energy diffusion, and the temporal dimension truncates redundant delay components, which not only eliminates the sparsity violation problem, but also significantly reduces the amount of data, thus solving the defects of low compression efficiency and heavy feedback burden of traditional schemes.

[0016] On the other hand, based on Joint Source Channel Coding (JSCC), encoding, noisy transmission and decoding are completed. The encoding adapts to the bandwidth requirements, the transmission matches the real channel noise, and the decoding accurately recovers the CSI. This breaks through the limitations of traditional coding separation, ensures the reconstruction accuracy in noisy environments, and finally feeds back the high-quality target CSI to the user end, providing support for high reliability and high-speed communication of multi-antenna systems, and making up for the shortcomings of poor robustness and insufficient accuracy of traditional solutions. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a channel state information feedback method provided in this embodiment of the disclosure; Figure 2 A schematic diagram of near-field communication in an XL-MIMO system model for a channel state information feedback method provided in a disclosed embodiment; Figure 3 A schematic diagram of the angle-distance domain sparse projection of the intermediate channel state information matrix of a channel state information feedback method provided in an embodiment of this disclosure; Figure 4 A grayscale schematic diagram of a specified channel state information matrix for a channel state information feedback method provided in an embodiment of this disclosure; Figure 5 An architecture diagram of the MF module in the decoding module of a channel state information feedback method provided in an embodiment of this disclosure; Figure 6 A schematic diagram illustrating the specific architecture of a channel state information feedback method provided in this embodiment of the disclosure; Figure 7 A schematic diagram showing the decoding processing comparison results of a channel state information feedback method provided in an embodiment of this disclosure; Figure 8 A schematic diagram of a channel state information feedback device provided in an embodiment of this disclosure; Figure 9 This is a schematic diagram of an electronic device provided in an embodiment of the present disclosure. Detailed Implementation

[0018] The present disclosure will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present disclosure and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the drawings, not the entire structure.

[0019] To enable those skilled in the art to better understand the technical solutions of this disclosure, and to fully understand and implement the process of how this disclosure applies technical means to solve technical problems and achieve corresponding technical effects, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, not all embodiments. The embodiments of this disclosure and the various features within them can be combined with each other without conflict, and the resulting technical solutions are all within the protection scope of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort should fall within the protection scope of this disclosure.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. 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 comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0022] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0023] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0024] Research has revealed that Channel State Information (CSI) is a channel attribute of a communication link. It describes the signal attenuation factor along each transmission path. CSI enables communication systems to adapt to current channel conditions, providing a guarantee for high-reliability, high-speed communication in multi-antenna systems.

[0025] In related technologies, during the channel state information feedback process, spherical waves cause the channel path to spread in the angular domain. This leads to a single near-field path component spreading to multiple angles, thereby disrupting the sparsity of the channel in the angular domain.

[0026] Based on the above research, this disclosure provides a feedback method for channel state information. Through spatial-temporal dual-dimensional compression, the spatial dimension relies on the angle-distance joint representation to capture the near-field CSI coupling characteristics and avoid energy diffusion. The temporal dimension truncates redundant delay components, which not only eliminates the sparsity destruction problem, but also significantly reduces the amount of data, thus solving the defects of low compression efficiency and heavy feedback burden of traditional schemes.

[0027] On the other hand, based on Joint Source Channel Coding (JSCC), encoding, noisy transmission and decoding are completed. The encoding adapts to the bandwidth requirements, the transmission matches the real channel noise, and the decoding accurately recovers the CSI. This breaks through the limitations of traditional coding separation, ensures the reconstruction accuracy in noisy environments, and finally feeds back the high-quality target CSI to the user end, providing support for high reliability and high-speed communication of multi-antenna systems, and making up for the shortcomings of poor robustness and insufficient accuracy of traditional solutions.

[0028] To facilitate understanding of this embodiment, a channel state information feedback method disclosed in this disclosure will first be described in detail. The execution subject of the channel state information feedback method provided in this disclosure is generally an electronic device with a certain computing capability. In some possible implementations, this data interaction method can be implemented by a processor calling computer-readable instructions stored in memory.

[0029] The embodiments of this disclosure can be applied to channel state information feedback in near-field environments (i.e., near-field communication), referring to... Figure 2The diagram shown illustrates near-field communication in an XL-MIMO system model of a channel state information feedback method provided in a public embodiment, including signal transmission status in the near field of multiple base stations (BS) and user equipment (USER) connected in communication.

[0030] This can be applied to downlink channel state information feedback XL-MIMO (Extremely Large-scale Massive Multiple-Input-Multiple-Output) systems in the near-field domain.

[0031] See Figure 1 The diagram shows a flowchart of a channel state information feedback method provided in an embodiment of this disclosure. The method includes steps S101 to S104, wherein: S101. Obtain the original channel state information matrix, and compress the original channel state information in both spatial and temporal dimensions to obtain the specified channel state information matrix.

[0032] In embodiments of this disclosure, firstly, communication signals received by the user equipment (UE) can be acquired. Secondly, a first initial channel state matrix can be determined based on the communication signals.

[0033] Here, the first initial channel state matrix H satisfies the following condition: ; Where y is the communication signal and v is the precoded carrier. N1 represents the number of antennas equipped at the base station (BS), and n represents the additive white Gaussian noise received by the user equipment. Here, N1 = N t N t The number of transmit antennas is preset for the base station.

[0034] Subsequently, the first initial channel state matrix can be processed based on the path components of the transmitting antenna under geometric free space propagation to determine the path components of each antenna of the base station, and the second initial channel state matrix can be determined based on the path components.

[0035] Here, the path component (i.e., channel) H(n2, n1) between the n1-th antenna of the base station and the n2-th antenna of the user equipment corresponding to the second initial channel state matrix satisfies the following condition: ; Where n1∈(1,2,...,N1), n2∈(1,2,...,N2), N2 is the number of user equipment antennas, N2=1. Let n be the transmission distance between the n1-th antenna of the base station and the n2-th antenna of the user equipment. This refers to the normalized free-space path loss of the base station antenna.

[0036] Specifically, the path components can be used to determine the different transmission paths that each base station's antenna pair experiences in the near-field domain of the XL-MMO (Extremely Large-scale Massive Multiple-Input-Multiple-Output) system.

[0037] Here, after determining the path components corresponding to each antenna of the base station, the second initial channel state matrix can be obtained based on the path components of each antenna.

[0038] Subsequently, the original channel state information matrix can be determined based on the second initial channel state matrix. The original channel state information matrix can be used to indicate the distance between the base station antenna and the user equipment antenna, the relative angle between the base station and the user equipment, and the signal transmission angle of the base station. The original channel state information matrix meets the following conditions: .

[0039] Here, after determining the original channel state information matrix, the original channel state information can be compressed sequentially in the spatial and temporal dimensions to obtain the specified channel state information matrix.

[0040] S102. The specified channel state information matrix is ​​encoded using joint source-channel coding technology to obtain the encoded channel state information matrix.

[0041] In the embodiments of this disclosure, a specified channel state information matrix can be encoded using the encoder module in the JSCC (Joint Source-Channel Coding) architecture to obtain the encoded channel state information matrix.

[0042] S103. Based on the channel quality status, Gaussian noise superposition is applied to the encoded channel information state matrix to obtain the superimposed channel state information matrix.

[0043] In the embodiments of this disclosure, Gaussian noise can be superimposed onto the encoded channel state information matrix by referring to the channel quality status of different channels to obtain the superimposed channel state information matrix.

[0044] S104. The superimposed channel state information matrix is ​​decoded using joint source-channel coding technology to obtain the target channel state information, and the target channel state information is fed back to the user terminal.

[0045] In the embodiments of this disclosure, the target channel state information matrix can be obtained by decoding the superimposed channel state information matrix through the RefineNet module in the JSCC (Joint Source-Channel Coding) architecture.

[0046] After determining the target channel state information matrix, the target channel state information matrix can be restored to obtain the target channel state information.

[0047] In the embodiments of this disclosure, firstly, the original channel state information matrix is ​​obtained, and the original channel state information is compressed in both spatial and temporal dimensions to obtain a specified channel state information matrix; secondly, the specified channel state information matrix is ​​encoded using joint source-channel coding technology to obtain an encoded channel state information matrix; thirdly, based on the channel quality state, the encoded channel state information matrix is ​​subjected to Gaussian noise superposition processing to obtain a superimposed channel state information matrix; finally, the superimposed channel state information matrix is ​​decoded using joint source-channel coding technology to obtain the target channel state information, and the target channel state information is fed back to the user terminal.

[0048] In the above implementation, through spatial-temporal dual-dimensional compression, the spatial dimension relies on the angle-distance joint characterization to capture the near-field CSI coupling characteristics and avoid energy diffusion, while the temporal dimension truncates redundant delay components, which not only eliminates the sparsity violation problem, but also significantly reduces the amount of data, thus solving the defects of low compression efficiency and heavy feedback burden of traditional schemes.

[0049] On the other hand, based on Joint Source Channel Coding (JSCC), encoding, noisy transmission and decoding are completed. The encoding adapts to the bandwidth requirements, the transmission matches the real channel noise, and the decoding accurately recovers the CSI. This breaks through the limitations of traditional coding separation, ensures the reconstruction accuracy in noisy environments, and finally feeds back the high-quality target CSI to the user end, providing support for high reliability and high-speed communication of multi-antenna systems, and making up for the shortcomings of poor robustness and insufficient accuracy of traditional solutions.

[0050] In one optional implementation, the original channel state information is compressed in both spatial and temporal dimensions to obtain a specified channel state information matrix, specifically including the following steps: First, the original channel state information matrix is ​​compressed in the spatial dimension to obtain the intermediate channel state information matrix. Then, the intermediate channel state information matrix is ​​compressed in the time dimension to obtain the specified channel state information matrix.

[0051] In the embodiments of this disclosure, the original channel state information matrix can be zero-filled and subjected to DFT (Discrete Fourier Transform), and the transformed matrix can be restored to the dimension of the original channel state information matrix to obtain the specified channel state information matrix.

[0052] Here, the near-field channel state information is affected by the curvature of the spherical wavefront, and the channel response depends on both distance and angle. The spatial phase term of the original channel state information matrix contains nonlinear components. The original channel state information matrix can be processed in the following way to obtain the intermediate channel state information matrix.

[0053] First, an angle-distance codebook is constructed to jointly represent the angle-distance dimension. Then, the spatial domain CSI is transformed to the angle-distance domain by codebook projection, so that the multipath components are sparsely concentrated at a specific angle-distance. Finally, the sparse matrix is ​​extracted to obtain the intermediate channel state information matrix.

[0054] Here, the original channel state information matrix is ​​compressed in the spatial dimension to obtain an intermediate channel state information matrix, including: Based on formula Determine the intermediate channel state information matrix; Where H is the original channel state information matrix. M is the number of antennas in the original channel, N is the number of subcarriers, and W... AD For angle-distance codebook, K AD These are the basis vectors bound by angle and distance.

[0055] Reference Figure 3 The diagram shown is a schematic diagram of the angle-distance domain sparse projection of the intermediate channel state information matrix of a channel state information feedback method provided in an embodiment of this disclosure, wherein: the horizontal axis is the angle-distance joint combination index (Polar Domain Index), and the vertical axis is the normalized energy (Normalized Energy).

[0056] Visualization (polar peak) and quantization metrics (NMSE) demonstrate that the sparse CSI matrix of the preprocessed output can both reduce dimensionality and retain core channel features.

[0057] like Figure 3As shown, the normalized mean square error (NMSE) between the original channel state information matrix and the intermediate channel state information matrix after codebook transformation, SOMP algorithm processing, and inverse codebook transformation is less than -20 dB. Therefore, the intermediate channel state information matrix after codebook transformation can be considered to be approximately the result after sparse processing by the SOMP algorithm. The number of channel paths L is 12, and there are exactly 12 peaks in the polar domain channel, demonstrating the sparsity of the intermediate channel state information matrix.

[0058] In the above implementation, a custom codebook W is used. AD Projecting the spatial domain CSI directly onto the angle-distance domain. Custom codebook W AD Corresponding to the physically bound angle-range propagation path, the multipath components exhibit sparsity in this domain. Extracting the main components and compressing them yields the intermediate channel state information matrix. In related technologies, multipath components are projected onto a discrete angle basis using 2D-DFT (2D Discrete Fourier Transform) to achieve efficient compression, leveraging the sparsity of the angle domain under the plane wave assumption. However, this approach may introduce energy leakage and disrupt the sparsity. This disclosure avoids the inability of traditional angle bases to capture range-dimensional coupling features through the aforementioned compression method.

[0059] Here, after determining the intermediate channel state information matrix, a DFT transformation can be performed on the frequency domain of the intermediate channel state information matrix to convert it into the time delay domain; then, the first Nc rows corresponding to the dominant time delay path can be extracted to obtain the specified channel state information matrix.

[0060] Since the multipath delay is concentrated within a finite time range, redundant rows with a value of 0 can be deleted, thus obtaining a specified channel state information matrix that retains the first Nc rows.

[0061] Here, the intermediate channel state information matrix is ​​compressed in the time dimension to obtain the specified channel state information matrix, including: Based on formula Determine the specified channel state information matrix; in, W is the intermediate channel state information matrix. TD The time-delay domain DFT transformation matrix, Nc represents the number of dominant delay paths.

[0062] Reference Figure 4 The image shown is a grayscale schematic diagram of a specified channel state information matrix in a channel state information feedback method provided in an embodiment of this disclosure.

[0063] In the above implementation, angle-distance domain compression and time-delay domain compression are combined into a complete formula, namely the formula for determining the specified channel state information matrix. Through the joint projection of codebook transformation and DFT transformation, the original channel state information matrix is ​​effectively compressed and the core physical features are preserved.

[0064] In an optional implementation, the specified channel state information matrix is ​​encoded using a joint source-channel coding technique to obtain the encoded channel state information matrix, specifically including the following steps: Based on formula Encoding is performed to obtain the encoded channel state information matrix; Among them, f en ( ) represents the compression function of the joint source-channel coding technique, θ en H represents the parameters of the neural network. s This specifies the channel state information matrix.

[0065] In the embodiments of this disclosure, a specified channel state information matrix can be encoded using the encoder module in the JSCC (Joint Source-Channel Coding) architecture to obtain the encoded channel state information matrix.

[0066] Here, as Figure 6 As shown, firstly, after the channel state information is preprocessed to obtain the specified channel state information matrix, the specified channel state information matrix is ​​input into the input module.

[0067] Secondly, the Encoder module processes the specified channel state information matrix to obtain the encoded channel state information matrix.

[0068] Here, the Encoder module includes a 3×3 convolutional submodule (Conv 3×3) for shallow feature extraction of a specified channel state information matrix. This size of convolutional kernel efficiently captures local feature patterns in the data, laying the foundation for subsequent deep processing.

[0069] Here, the Encoder module also includes an MF sub-module (MF Layer). The data processed by Conv 3×3 enters the MF (Multi-scale Feature) sub-module. As the core unit of feature processing, the MF sub-module aims to integrate multi-scale contextual information and enhance high-frequency details. In practical applications, features at different scales correspond to different levels of information in the data. Small-scale features often contain rich details, while large-scale features reflect the overall structure and contextual relationships. By organically combining these different scales of information, it ensures that both the details and the overall structure of the data are fully expressed.

[0070] Here, the Encoder module also includes a 5×5 convolutional submodule (i.e., Conv 5×5). This 5×5 convolutional submodule further processes the data output by the MFLayer. Its larger kernel size allows it to extract deeper features and uncover more complex and abstract patterns in the data. After these convolutional operations, the resulting feature map is flattened and compressed to the complex domain through a fully connected layer. The use of complex representation here has profound physical significance. In wireless communication, IQ signals are a common signal form, and their physical characteristics determine the natural suitability of complex representation. The real and imaginary parts of the complex number correspond to the in-phase and quadrature components of orthogonal carriers, respectively. This representation perfectly matches the RF modulation and demodulation process, enabling seamless integration with subsequent wireless transmission stages. After completing the complex representation, the complex vector also needs to undergo strict power constraints. This is to meet the power limitations of actual channel transmission, ensuring that the signal transmission in the channel will not cause interference due to excessive power, nor will it lead to signal quality degradation due to insufficient power.

[0071] Here, the Encoder module also includes a fully connected layer submodule (i.e., FC) for performing feature dimensionality transformation and compression on the Conv 5×5 output data.

[0072] Here, the Encoder module also includes a complex submodule (i.e., Complex Layer), which is used to split the data output by FC into "real part-imaginary part" to construct a complex vector, ensuring that the power limit of channel transmission is met and avoiding signal distortion.

[0073] In one optional implementation, the encoded channel state information matrix is ​​subjected to Gaussian noise superposition processing according to the channel quality state to obtain the superimposed channel state information matrix, specifically including the following steps: First, obtain the signal-to-noise ratio (SNR) parameter; Secondly, based on the channel quality status and the signal-to-noise ratio parameter, the target Gaussian noise of the specified channel state information matrix is ​​determined; Finally, the channel state information matrix is ​​encoded from the superimposed values ​​of the target Gaussian noise to obtain the superimposed channel state information matrix.

[0074] In the embodiments of this disclosure, a pre-configured signal-to-noise ratio (SNR) parameter can be obtained. The SNR parameter simulates different channel quality conditions. In actual communication scenarios, channels are subject to various noise interferences. By setting different SNR parameters, various transmission environments, ranging from ideal channels to poorly configured channels, can be simulated.

[0075] In the above implementation, the performance of the JSCC network under different conditions can be comprehensively tested by simulating various transmission environments, from ideal channels to adverse channels.

[0076] like Figure 6 As shown, the core output module (Codeword1) in the encoding stage is used to receive the encoded channel state information matrix, so that Gaussian noise can be superimposed on the encoded channel state information matrix.

[0077] The core input module (Codeword2) in the encoding stage is used to transmit the superimposed channel state information matrix to the decoding module.

[0078] In an optional implementation, the superimposed channel state information matrix is ​​decoded using joint source-channel coding techniques to obtain the target channel state information, specifically including the following steps: Multi-scale feature enhancement processing is performed on the superimposed channel state information matrix to obtain the first superimposed matrix; The first superimposed matrix is ​​subjected to progressive convolution to obtain the second superimposed matrix; The target channel state information matrix is ​​obtained by processing the second superimposed matrix using an activation function. The target channel state information matrix is ​​restored to obtain the target channel state information.

[0079] In embodiments of this disclosure, it can be based on formulas The superimposed channel state information matrix is ​​decoded to obtain the target channel state information matrix.

[0080] Among them, f de ( ) is the decompression function, θ de These are the neural network parameters for the decoder.

[0081] Here, the mean squared error can be used as the loss function for both encoding and decoding. Substituting the encoding and decoding formulas into the loss function yields the optimized neural network model: ; in, f en The optimal parameters for ( ) f de The optimal parameters for ().

[0082] like Figure 6 As shown, the decoding module includes a complex layer submodule, a fully connected layer submodule, a 5×5 convolutional submodule, an MF layer submodule, a RefineNet module, and a 3×3 convolutional submodule.

[0083] The RefineNet module includes a progressive receptive field expansion structure consisting of 3×3 convolutional layers (Conv 3×3), 5×5 convolutional layers (Conv 5×5), and 9×9 convolutional layers (Conv 9×9).

[0084] First, the fully connected layer submodule remaps the complex signal into a spatial feature map, realizing the transformation from a one-dimensional complex vector to a two-dimensional feature map, which provides a foundation for subsequent feature recovery and enhancement.

[0085] Then, the 5×5 convolutional submodule comes into play again, restoring the local features of the data and allowing the detailed information in the feature map to be reconstructed.

[0086] Afterward, the data enters the MF module for a second multi-scale feature enhancement to further optimize the quality of the feature maps and ensure that the multi-scale information of the data can be fully utilized.

[0087] Next, the RefineNet module processes the data using a highly innovative residual connection design. Each RefineNet module contains a progressively expanding receptive field structure composed of 3×3, 5×5, and 9×9 convolutions. Large convolutional kernels, such as 9×9 kernels, effectively capture long-range dependencies in the data, which is crucial for processing data with complex structures. The introduction of residual connections effectively alleviates the vanishing gradient problem common in deep neural networks, ensuring smooth parameter updates during training and improving the network's training efficiency and stability.

[0088] Finally, after processing by the RefineNet module, the reconstruction result is output by a 3×3 convolution combined with the Sigmoid activation function.

[0089] Among them, the MF module, as a key component of the JSCC system architecture, is the core of feature enhancement through its unique multi-branch architecture. The design philosophy of this module is to simultaneously capture high-frequency details and global context information to meet diverse data processing needs.

[0090] Here, after determining the target channel state information matrix, post-processing (i.e., restoration) can be performed on it. First, zero values ​​can be added to the target channel state information matrix according to its dimensions to restore the redundant components truncated in the preprocessing stage. Then, an inverse DFT transformation can be performed on the zero-padded matrix to restore the time delay domain to the empty frequency domain, ultimately obtaining the target channel state information.

[0091] Reference Figure 5 The diagram shown is an architecture diagram of the MF module in the decoding module of a channel state information feedback method provided in this embodiment of the present disclosure, including: a fusion submodule (Fusion Layer 1), a 1×1 convolution submodule (Conv 1×1), a 3×3 convolution submodule (Conv 3×3), a 5×5 convolution submodule (Conv 5×5), a fusion submodule (Fusion Layer 2), a 7×7 convolution submodule (Conv 7×7), a weighted fusion submodule (Weighted Fusion Layer), and an output submodule (Output).

[0092] Here, the input features are first concatenated along the channel dimension. The high-frequency feature submodule (HighFreqBranch), the context feature submodule (ContextBranch), and the original feature submodule (IdentityBranch) are fused together through the fusion submodule (Fusion Layer1) to form a rich and comprehensive feature combination. This concatenation method allows different types of information to interact in the same space, providing ample material for subsequent feature extraction.

[0093] Subsequently, 1×1, 3×3, and 5×5 convolutional operations are applied to the concatenated features using 1×1 (Conv 1×1), 3×3 (Conv 3×3), and 5×5 (Conv 5×5) convolutional submodules, respectively. These convolutional kernels of different sizes can extract contextual information of different spatial ranges. 1×1 convolution can adjust the channel dimension and integrate features without changing the spatial size; 3×3 convolution is suitable for capturing medium-scale local features; and 5×5 convolution can obtain a wider range of contextual information.

[0094] After extracting multi-scale features, the features are fused using the Fusion Layer 2 module, and then processed by the fused features using the 7×7 Convolutional Layer 2 module. This module uses 7×7 convolutions to generate a spatial weight map. The large-size convolutional kernels enable it to effectively model long-distance dependencies between pixels and accurately capture the correlations between different regions in the data.

[0095] Next, the data processed by the 7×7 convolutional submodule (Conv 7×7) is processed by the weighted fusion submodule (Weighted Fusion Layer). Here, the weight values ​​are normalized to the 0-1 range to obtain reasonable spatial attention weights. Multi-scale features are then weighted and fused based on these weights. Compared to simple addition or concatenation fusion methods, this spatial attention-based weighted fusion design effectively solves the problem of limited receptive field in single-path convolution. It gives the network the ability to adaptively focus on the optimal feature scale in different regions, especially in high-noise channel environments, accurately suppressing interference from irrelevant regions and highlighting the response of important features, thereby significantly improving the accuracy and robustness of data processing.

[0096] Finally, the target channel state information matrix obtained after processing by the Weighted Fusion Layer is output through the Output layer.

[0097] Reference Figure 6 The diagram shown is a schematic representation of the specific architecture of a channel state information feedback method provided in this embodiment of the present disclosure, including: an input module, an Encoder module, and a Decoder module.

[0098] The Encoder module includes: a 3×3 convolutional submodule (Conv 3×3), an MF submodule (MF Layer), a 5×5 convolutional submodule (i.e., Conv 5×5), a fully connected layer submodule (i.e., FC), and a complex layer submodule (i.e., ComplexLayer).

[0099] The Decoder module includes: a complex layer submodule, a fully connected layer submodule, a 5×5 convolutional submodule, an MF layer submodule, a RefineNet module, and a 3×3 convolutional submodule.

[0100] Here, the Encoder module and the Decoder module include the core output module (Codeword1) and the core input module (Codeword2) of the encoding stage.

[0101] The core output module (Codeword1) in the encoding stage is used to receive the encoded channel state information matrix output by the complex sub-module (i.e., Complex Layer) in the Encoder module, so that Gaussian noise can be superimposed on the encoded channel state information matrix.

[0102] The core input module (Codeword2) in the encoding stage is used to transmit the superimposed channel state information matrix to the decoding module.

[0103] Reference Figure 7 The diagram shown is a comparison of decoding processing results of a channel state information feedback method provided in this embodiment of the present disclosure, wherein: Ablation experiments were conducted on the MF module and other forms of multi-scale fusion networks. The results show that the multi-scale fusion method used by MR outperforms the NMSE of addition, multiplication and splicing networks under different SNR environments.

[0104] In the above embodiments, element-wise addition of the three features can lead to mutual interference between different semantic features; element-wise multiplication of the three features can excessively suppress non-overlapping activation regions; and using 1×1 convolution for dimensionality reduction after concatenating the three features can result in a lack of spatial adaptability. The multi-scale fusion method used in this disclosure enhances feature extraction capabilities by extracting diverse features and introducing dynamic weighted fusion with spatial attention.

[0105] Based on the same inventive concept, this disclosure also provides a channel state information feedback device corresponding to the channel state information feedback method. Since the principle of the device in this disclosure for solving the problem is similar to the channel state information feedback method described above in this disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0106] Reference Figure 8 The diagram shown is a schematic of a channel state information feedback device provided in an embodiment of this disclosure. The device includes: a compression unit 81, an encoding unit 82, a noise superposition unit 83, and a decoding unit 84; wherein: A compression unit is used to acquire the original channel state information matrix and perform spatial and temporal compression on the original channel state information to obtain a specified channel state information matrix. The encoding unit is used to encode the specified channel state information matrix using joint source-channel coding techniques to obtain the encoded channel state information matrix. The noise superposition unit is used to perform Gaussian noise superposition processing on the encoded channel information state matrix according to the channel quality state to obtain the superimposed channel state information matrix. The decoding unit is used to decode the superimposed channel state information matrix using the joint source channel coding technique to obtain the target channel state information, and to feed the target channel state information back to the user terminal.

[0107] The processing flow of each module in the device and the interaction flow between each module can be referred to the relevant descriptions in the above method embodiments, and will not be detailed here.

[0108] Corresponding to Figure 1 In addition to the method for feeding back channel state information, this disclosure also provides an electronic device 900, such as... Figure 9 The diagram shown is a structural schematic of an electronic device 900 provided in an embodiment of this disclosure, including: The system includes a processor 91, a memory 92, and a bus 93. The memory 92 stores execution instructions and includes main memory 921 and external memory 922. The main memory 921, also called internal memory, temporarily stores the computational data in the processor 91, as well as data exchanged with external memory such as a hard disk. The processor 91 exchanges data with the external memory 922 through the main memory 921. When the electronic device 900 is running, the processor 91 communicates with the memory 92 through the bus 93, causing the processor 91 to execute the following instructions: Obtain the original channel state information matrix, and compress the original channel state information in both spatial and temporal dimensions to obtain the specified channel state information matrix; The specified channel state information matrix is ​​encoded using joint source-channel coding techniques to obtain the encoded channel state information matrix. Based on the channel quality status, the encoded channel information state matrix is ​​subjected to Gaussian noise superposition processing to obtain the superimposed channel state information matrix. The superimposed channel state information matrix is ​​decoded using the joint source-channel coding technique to obtain the target channel state information, which is then fed back to the user terminal.

[0109] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0110] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0111] Additionally, as used herein, the "or" used in a list of items beginning with "at least one" indicates a separate list, such that a list of, for example, "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not imply that the described example is preferred or better than other examples.

[0112] It should also be noted that in the systems and methods of this disclosure, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0113] Various changes, substitutions, and modifications can be made to the technology described herein without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of this disclosure is not limited to the specific aspects of the processes, machines, manufactures, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufactures, events, means, methods, or actions that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Therefore, the appended claims include such processes, machines, manufactures, events, means, methods, or actions within their scope.

[0114] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0115] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A method for feedback of channel state information, characterized in that, include: Obtain the original channel state information matrix, and compress the original channel state information in both spatial and temporal dimensions to obtain the specified channel state information matrix; The specified channel state information matrix is ​​encoded using joint source-channel coding techniques to obtain the encoded channel state information matrix. Based on the channel quality status, the encoded channel information state matrix is ​​subjected to Gaussian noise superposition processing to obtain the superimposed channel state information matrix. The superimposed channel state information matrix is ​​decoded using the joint source-channel coding technique to obtain the target channel state information, which is then fed back to the user terminal.

2. The method according to claim 1, characterized in that, The step of compressing the original channel state information in both spatial and temporal dimensions to obtain a specified channel state information matrix includes: The original channel state information matrix is ​​compressed in the spatial dimension to obtain an intermediate channel state information matrix; The intermediate channel state information matrix is ​​compressed in the time dimension to obtain the specified channel state information matrix.

3. The method according to claim 2, characterized in that, The step of compressing the original channel state information matrix in the spatial dimension to obtain an intermediate channel state information matrix includes: Based on formula Determine the intermediate channel state information matrix; Wherein, H is the original channel state information matrix. M is the number of antennas in the original channel, N is the number of subcarriers, and W... AD For angle-distance codebook, K AD These are the basis vectors bound by angle and distance.

4. The method according to claim 3, characterized in that, The step of compressing the intermediate channel state information matrix in the time dimension to obtain the specified channel state information matrix includes: Based on formula Determine the specified channel state information matrix; in, W is the intermediate channel state information matrix. TD The time-delay domain DFT transformation matrix, Nc represents the number of dominant delay paths.

5. The method according to claim 1, characterized in that, The step of encoding the specified channel state information matrix using joint source-channel coding technology to obtain the encoded channel state information matrix includes: Based on formula Encoding is performed to obtain the encoded channel state information matrix; Among them, f en ( ) represents the compression function of the joint source-channel coding technique, θ en H represents the parameters of the neural network. s This specifies the channel state information matrix.

6. The method according to claim 1, characterized in that, The step of performing Gaussian noise superposition processing on the encoded channel state information matrix according to the channel quality state to obtain the superimposed channel state information matrix includes: Obtain the signal-to-noise ratio parameter; Based on the channel quality state and the signal-to-noise ratio parameter, determine the target Gaussian noise of the specified channel state information matrix; The target Gaussian noise is superimposed onto the encoded channel state information matrix to obtain the superimposed channel state information matrix.

7. The method according to claim 1, characterized in that, The step of decoding the superimposed channel state information matrix using the joint source-channel coding technique to obtain the target channel state information includes: The superimposed channel state information matrix is ​​subjected to multi-scale feature enhancement processing to obtain the first superimposed matrix; The first superimposed matrix is ​​subjected to progressive convolution to obtain the second superimposed matrix; The target channel state information matrix is ​​obtained by processing the second superimposed matrix using an activation function. The target channel state information matrix is ​​restored to obtain the target channel state information.

8. A channel state information feedback device, characterized in that, include: A compression unit is used to acquire the original channel state information matrix and perform spatial and temporal compression on the original channel state information to obtain a specified channel state information matrix. The encoding unit is used to encode the specified channel state information matrix using joint source-channel coding techniques to obtain the encoded channel state information matrix. The noise superposition unit is used to perform Gaussian noise superposition processing on the encoded channel information state matrix according to the channel quality state to obtain the superimposed channel state information matrix. The decoding unit is used to decode the superimposed channel state information matrix using the joint source channel coding technique to obtain the target channel state information, and to feed the target channel state information back to the user terminal.

9. A non-transitory computer-readable storage medium, characterized in that, The storage medium stores at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the method as described in any one of claims 1-7.

10. An electronic device, characterized in that, Includes a processor and the non-transitory computer-readable storage medium of claim 9.

Images