Channel estimation method and device for wireless communication, electronic equipment and storage medium

Abstract

The present application relates to the field of wireless communication, and especially relates to a channel estimation method and device for wireless communication, electronic equipment and storage medium, the method comprising: extracting a first channel feature matrix in each scale and each direction from an initial channel state matrix contained in pilot information; inputting the latest first channel feature matrix into a denoising network to obtain a second channel feature matrix; correcting the second channel feature matrix according to a noise error value between the first channel feature matrix and the second channel feature matrix to obtain the latest first channel feature matrix; and continuing to denoise and correct until a preset denoising number is reached; and determining a target channel state matrix corresponding to the pilot information based on all the final obtained first channel feature matrices. The present application performs channel estimation in a manner of removing noise and correction through multiple iterations, which can guarantee channel estimation performance and improve robustness in the case of high noise level or large channel condition change.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wireless communication, in particular to a channel estimation method and device for wireless communication, an electronic device and a storage medium. BACKGROUND

[0002] In a wireless communication system, channel estimation is one of the key technologies to guarantee communication quality, and its essence is to accurately model and estimate the transmission characteristics of a wireless channel by means of signal processing technology. The wireless channel is easily affected by factors such as multipath fading, Doppler effect and noise interference, which makes it essential for the receiving end to accurately estimate the channel state information to ensure effective demodulation and recovery of the signal.

[0003] Traditional channel estimation methods often use end-to-end neural networks, with pilot information as input and channel estimation results as output.

[0004] However, in the case of high noise level or large channel condition change, the evaluation performance of the traditional method will decrease significantly, and the robustness is insufficient. SUMMARY

[0005] Therefore, the purpose of the present application is to provide a channel estimation method and device for wireless communication, an electronic device and a storage medium, which can guarantee channel estimation performance and improve robustness in the case of high noise level or large channel condition change.

[0006] In a first aspect, the embodiments of the present application provide a channel estimation method for wireless communication, comprising:

[0007] extracting a first channel feature matrix in each scale and each direction from an initial channel state matrix contained in pilot information;

[0008] for each first channel feature matrix, inputting the latest first channel feature matrix into a denoising network to obtain a second channel feature matrix; the denoising network is obtained by training an actual channel state matrix and a complete noise channel feature matrix corresponding to the actual channel state matrix obtained by using a noise scheduling parameter; the noise scheduling parameter is used to control the noise addition ratio;

[0009] correcting the second channel feature matrix according to the noise error value between the first channel feature matrix and the second channel feature matrix to obtain the latest first channel feature matrix; and continuing to input the latest first channel feature matrix into the denoising network until a preset denoising number is reached;

[0010] determining a target channel state matrix corresponding to the pilot information based on all the final first channel feature matrices obtained.

[0011] In a possible implementation, the correcting the second channel feature matrix according to the noise error value between the first channel feature matrix and the second channel feature matrix to obtain the latest first channel feature matrix comprises:

[0012] determining a noise error adjustment strength according to a numerical relationship between the noise error value and a noise error threshold value;

[0013] correcting the second channel feature matrix according to the noise error adjustment strength and the noise error value to obtain the latest first channel feature matrix.

[0014] In a possible implementation, the determining the noise error adjustment strength according to the numerical relationship between the noise error value and the noise error threshold value comprises:

[0015] determining whether the noise error value is greater than the noise error threshold value;

[0016] if the noise error value is greater than the noise error threshold value, calculating the noise error adjustment strength according to the noise error value;

[0017] if the noise error value is less than or equal to the noise error threshold value, setting the noise error adjustment strength to 1.

[0018] In a possible implementation, the calculating the noise error adjustment strength according to the noise error value comprises:

[0019] substituting the noise error value into the following formula to obtain the noise error adjustment strength;

[0020] ;

[0021] wherein, the noise error adjustment strength is denoted as K, the noise error value is denoted as e, and the remaining processing times is denoted as n.

[0022] In a possible implementation, the correcting the second channel feature matrix according to the noise error adjustment strength and the noise error value to obtain the latest first channel feature matrix comprises:

[0023] substituting the noise error adjustment strength and the noise error value into the following formula to obtain the latest first channel feature matrix;

[0024] ;

[0025] wherein, the latest first channel feature matrix is denoted as H, a second channel feature matrix, a remaining processing number, a denoising network, a trainable parameter of the denoising network, a latest first channel feature matrix obtained when the remaining processing number is t+1.

[0026] In a possible implementation, the denoising network is trained by the following steps:

[0027] extracting, from an actual channel state matrix, an actual channel feature matrix of channel state information in each scale and each direction;

[0028] adding noise to the actual channel feature matrix according to the noise scheduling parameter to obtain a completely noisy channel feature matrix;

[0029] performing denoising on the completely noisy channel feature matrix by the denoising network and performing correction to obtain a restored channel state matrix corresponding to the completely noisy channel feature matrix;

[0030] updating parameters in the denoising network with the least difference between the restored channel state matrix and the actual channel state matrix as an optimization target, and continuing to perform denoising on the completely noisy channel feature matrix by the denoising network and performing correction until a preset iteration stopping condition is met.

[0031] In a possible implementation, the adding noise to the actual channel feature matrix according to the noise scheduling parameter to obtain a completely noisy channel feature matrix comprises:

[0032] inputting the noise scheduling parameter and the actual channel feature matrix into the following formula to obtain a completely noisy channel feature matrix;

[0033] ;

[0034] ;

[0035] wherein, the completely noisy channel feature matrix, the noise scheduling parameter of the i th iteration, a preset number of iterations of adding noise, the noise scheduling parameter of the j+1 th iteration, a standard complex Gaussian noise value of the j th iteration that conforms to a standard normal distribution I is a covariance matrix of the standard normal distribution, the actual channel feature matrix.

[0036] In a second aspect, the embodiments of the present application further provide a channel estimation device for wireless communication, the device comprising:

[0037] an extraction module configured to extract, from an initial channel state matrix contained in pilot information, a first channel feature matrix in each scale and each direction;

[0038] an input module configured to input, for each first channel feature matrix, a latest first channel feature matrix into a denoising network to obtain a second channel feature matrix, wherein the denoising network is trained based on an actual channel state matrix and a complete noise channel feature matrix corresponding to the actual channel state matrix obtained by using a noise scheduling parameter, and the noise scheduling parameter is used to control a noise addition ratio;

[0039] a correction module configured to correct the second channel feature matrix according to a noise error value between the first channel feature matrix and the second channel feature matrix to obtain a latest first channel feature matrix, and continue to input the latest first channel feature matrix into the denoising network until a preset denoising number is reached;

[0040] a determination module configured to determine a target channel state matrix corresponding to the pilot information based on all the finally obtained first channel feature matrices.

[0041] In a possible implementation, the correction module is specifically configured to determine a noise error adjustment strength according to a numerical relationship between the noise error value and a noise error threshold value, and correct the second channel feature matrix according to the noise error adjustment strength and the noise error value to obtain a latest first channel feature matrix.

[0042] In a possible implementation, the correction module is further configured to:

[0043] determine whether the noise error value is greater than the noise error threshold value;

[0044] if the noise error value is greater than the noise error threshold value, calculate a noise error adjustment strength according to the noise error value;

[0045] if the noise error value is less than or equal to the noise error threshold value, set the noise error adjustment strength to 1.

[0046] In a possible implementation, the correction module is further configured to:

[0047] substitute the noise error value into the following formula to obtain the noise error adjustment strength;

[0048] ;

[0049] wherein, Adjust the intensity for noise error. This is the noise error value. This represents the remaining number of processing attempts.

[0050] In one possible implementation, the correction module is further configured to:

[0051] Substituting the noise error adjustment intensity and the noise error value into the following formula, the latest first channel feature matrix is ​​obtained;

[0052] ;

[0053] in, This is the latest first channel feature matrix. This is the second channel feature matrix. The remaining number of processing attempts. For noise reduction networks, These are the trainable parameters for the denoising network. This is the latest first channel feature matrix obtained when the remaining processing times are t+1.

[0054] In one possible implementation, the device further includes a training module, specifically configured to: extract actual channel feature matrices of channel state information at each scale and in each direction from the actual channel state matrix; iteratively add noise to the actual channel feature matrices according to the noise scheduling parameters to obtain a completely noisy channel feature matrix; denoise and correct the completely noisy channel feature matrix through the denoising network to obtain a recovered channel state matrix corresponding to the completely noisy channel feature matrix; update the parameters in the denoising network with the goal of minimizing the difference between the recovered channel state matrix and the actual channel state matrix; and continue to denoise and correct the completely noisy channel feature matrix through the denoising network until a preset iteration stop condition is met.

[0055] In one possible implementation, the training module is further configured to:

[0056] Input the noise scheduling parameters and the actual channel feature matrix into the following formula to obtain the complete noise channel feature matrix;

[0057] ;

[0058] ;

[0059] in, This is the complete noise channel feature matrix. Let be the noise scheduling parameters for the i-th iteration. The number of times noise is added to the preset iteration. Let J be the noise scheduling parameters for the (j+1)th iteration. The standard complex Gaussian noise value following a standard normal distribution in the j-th iteration. Let I be the covariance matrix of the standard normal distribution. This is the actual channel feature matrix.

[0060] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the channel estimation method for wireless communication as described in any of the first aspects.

[0061] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the channel estimation method for wireless communication as described in any of the first aspects.

[0062] This application provides a channel estimation method, apparatus, electronic device, and storage medium for wireless communication. The method includes: extracting a first channel feature matrix for each scale and direction from an initial channel state matrix contained in pilot information; for each first channel feature matrix, inputting the latest first channel feature matrix into a denoising network to obtain a second channel feature matrix; correcting the second channel feature matrix based on the noise error value between the first and second channel feature matrices to obtain a latest first channel feature matrix; and continuing to input the latest first channel feature matrix into the denoising network until a preset number of denoising iterations is reached; and determining the target channel state matrix corresponding to the pilot information based on all the finally obtained first channel feature matrices. This application performs channel estimation by iteratively removing and correcting noise, which can ensure channel estimation performance and improve robustness even under high noise levels or large changes in channel conditions. Attached Figure Description

[0063] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 A flowchart of a channel estimation method for wireless communication provided in an embodiment of this application is shown;

[0065] Figure 2 A flowchart illustrating the training process of the denoising network provided in an embodiment of this application is shown.

[0066] Figure 3 This illustration shows a schematic diagram of the structure of a channel estimation device for wireless communication provided in an embodiment of this application;

[0067] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0069] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0070] To enable those skilled in the art to utilize the content of this application, and in conjunction with the specific application scenario of "wireless communication," the following implementation methods are provided. For those skilled in the art, the general principles defined herein can be applied to other embodiments and application scenarios without departing from the spirit and scope of this application. Although this application is primarily described within the "field of wireless communication," it should be understood that this is merely an exemplary embodiment.

[0071] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0072] The following is a detailed description of a channel estimation method for wireless communication provided by an embodiment of this application.

[0073] Reference Figure 1 The diagram shown is a flowchart illustrating a channel estimation method for wireless communication provided in an embodiment of this application. The exemplary steps of this embodiment are described below:

[0074] S101. Extract the first channel feature matrix for each scale and each direction from the initial channel state matrix contained in the pilot information.

[0075] In the embodiments of this application, in a MIMO (Multiple-Input Multiple-Output) system, the initial channel state matrix (CSI, Where C is the set of complex numbers, and its size is . This describes the initial channel state between each antenna at the transmitting end and each antenna at the receiving end. Initial Channel State Matrix It is a complex matrix. In the Cartesian coordinate system, the distance from the Cartesian coordinate corresponding to each complex element in the initial channel state matrix to the origin represents the channel strength, and the angle between the Cartesian coordinate corresponding to each complex element and the origin represents the channel delay. The real part matrix ; , is the imaginary unit.

[0076] This application embodiment uses two-dimensional discrete wavelet transform (2D-DWT) to transform the real part matrix and the... It is decomposed into sub-bands of different scales and directions, namely the first channel feature matrix of different scales and directions.

[0077] Here, the first channel feature matrix includes an approximation coefficient matrix, a vertical detail coefficient matrix, a horizontal detail coefficient matrix, and a diagonal detail coefficient matrix. (1) The approximation coefficient matrix is ​​used to represent the low-frequency part of the channel state matrix, i.e., the smoothness or overall trend of the channel. They capture the main features of the channel but remove the detailed information. (2) The vertical detail coefficient matrix is ​​used to represent the high-frequency part of the channel state matrix in the vertical direction, i.e., the rapid changes or edge information of the channel in the vertical direction. They capture the detailed features of the channel in the vertical direction. (3) The horizontal detail coefficient matrix is ​​used to represent the high-frequency part of the channel state matrix in the horizontal direction, i.e., the rapid changes or edge information of the channel in the horizontal direction. They capture the detailed features of the channel in the horizontal direction. (4) The diagonal detail coefficient matrix is ​​used to represent the high-frequency part of the channel state matrix in the diagonal direction, i.e., the rapid changes or edge information of the channel in the diagonal direction. They capture the detailed features of the channel in the diagonal direction.

[0078] Specifically, the real part matrix or... is obtained by performing the two-dimensional discrete wavelet transform through the following steps. Decomposed into first channel feature matrices of different scales and directions:

[0079] First, for the real part matrix or each line Perform row filtering to obtain intermediate approximation coefficients. and intermediate detail coefficient :

[0080]

[0081] in, The number of rows in the real or imaginary part of the matrix. For real part matrix or The number of columns, intermediate approximation coefficients The value in the m-th row and k-th column of the array. For real part matrix or , For intermediate detail coefficients The value in the m-th row and k-th column of the array. For the selected wavelet-correlated low-pass filter coefficients, The high-pass filter coefficients associated with the selected wavelet.

[0082] Then, for the intermediate approximation coefficients and intermediate detail coefficient Each column Applying one-dimensional DWT, the first channel feature matrix at different scales and directions is obtained (its size is...). ):

[0083]

[0084] in, This is an approximate coefficient matrix. This is the vertical detail coefficient matrix. This is the horizontal detail coefficient matrix. This is the diagonal detail coefficient matrix.

[0085] S102. For each first channel feature matrix, input the latest first channel feature matrix into the denoising network to obtain the second channel feature matrix.

[0086] The denoising network is trained based on the actual channel state matrix and the complete noise channel feature matrix corresponding to the actual channel state matrix obtained by using noise scheduling parameters; the noise scheduling parameters are used to control the noise addition ratio.

[0087] Reference Figure 2The diagram shown is a flowchart of the training process for the denoising network provided in this application embodiment. The exemplary steps of this application embodiment are described below:

[0088] S201. Extract the channel state information from the actual channel state matrix, and extract the actual channel feature matrix at each scale and in each direction.

[0089] In the embodiments of this application, the method of extracting the actual channel feature matrix from the actual channel state matrix is ​​the same as the process of extracting the first channel feature matrix from the initial channel state matrix, and will not be described again here.

[0090] The actual channel state matrix contains the actual state of the channel.

[0091] S202. Based on the noise scheduling parameters, noise is iteratively added to the actual channel feature matrix to obtain the complete noise channel feature matrix.

[0092] In the embodiments of this application, during the forward process (training phase) of the diffusion model, Gaussian noise is gradually added to the inter-channel feature matrix until it becomes completely random noise, thus obtaining the complete noise channel feature matrix.

[0093] Here, in this process, the diffusion model simulates the degradation of data by progressively adding noise until the data becomes entirely noise. This process can be viewed as constructing a Markov chain, where each state depends only on the previous state. Since the true distribution is unknown, we use the state probabilities predicted by the diffusion model. To approximate a true state overview .

[0094] ;

[0095] ;

[0096] in, This represents the total number of diffusion steps (typically 200-1000 steps, controlling the noise addition rhythm; it can also be referred to as the preset number of iterations to add noise). Let be the probability.

[0097] In order to To obtain the Gaussian white noise distribution, we used two specific scalars. (Hyperparameters of the diffusion model) and , These are the trainable parameters of the denoising network in the diffusion model.

[0098] The calculation formula is:

[0099] ;

[0100] The following formula can be obtained by accumulating and superimposing the above calculation formulas:

[0101] ;

[0102] ;

[0103] in, This is the complete noise channel feature matrix. Let be the noise scheduling parameters for the i-th iteration. The number of times noise is added to the preset iteration. Let J be the noise scheduling parameters for the (j+1)th iteration. The standard complex Gaussian noise value following a standard normal distribution in the j-th iteration. Let I be the covariance matrix of the standard normal distribution. This is the actual channel feature matrix.

[0104] S203. After denoising the complete noisy channel feature matrix using a denoising network and then correcting it, the recovered channel state matrix corresponding to the complete noisy channel feature matrix is ​​obtained.

[0105] In this embodiment, for each fully noisy channel feature matrix, the latest fully noisy channel feature matrix is ​​input into the denoising network to obtain a third channel feature matrix; based on the noise error value between the latest fully noisy channel feature matrix and the third channel feature matrix, the third channel feature matrix is ​​corrected to obtain the latest fully noisy channel feature matrix; and the latest fully noisy channel feature matrix is ​​continued to be input into the denoising network until a preset number of denoising times is reached, after which the latest fully noisy channel feature matrix is ​​determined as the recovered channel state matrix.

[0106] S204. With the goal of minimizing the difference between the recovered channel state matrix and the actual channel state matrix, update the parameters in the denoising network; and continue to denoise and correct the completely noisy channel feature matrix through the denoising network until the preset iteration stopping condition is met.

[0107] In the embodiments of this application, during the forward pass of the diffusion model, a denoising network is used to learn how to predict noise at any time and optimize the relevant trainable parameters of the diffusion model (including the parameters in the denoising network).

[0108] Here, existing solutions generally use end-to-end neural networks, while this technical solution only uses an end-to-end denoising network to fit the noise (S201 to S204) and uses a diffusion model framework to solve the channel estimation problem, thus making the entire channel estimation method robust and adaptable to changing environmental information.

[0109] In addition, during the forward pass, CNNs can be used to fit the noise, such as fully connected networks or U-net, or the same effect can be achieved by increasing or decreasing the number of network layers.

[0110] Other generative networks such as GANs and VAEs can achieve similar results.

[0111] S103. Based on the noise error value between the first channel feature matrix and the second channel feature matrix, the second channel feature matrix is ​​corrected to obtain the latest first channel feature matrix; and the latest first channel feature matrix is ​​continued to be input into the denoising network until the preset number of denoising times is reached.

[0112] In this embodiment, the noise error adjustment intensity is determined based on the numerical relationship between the noise error value and the noise error threshold; the second channel feature matrix is ​​corrected based on the noise error adjustment intensity and the noise error value to obtain the latest first channel feature matrix.

[0113] Specifically, the noise error adjustment intensity is determined based on the numerical relationship between the noise error value and the noise error threshold, including:

[0114] Step 1: Determine whether the noise error value is greater than the noise error threshold.

[0115] In this embodiment, the noise error threshold is used to distinguish whether the noise in the latest first channel feature matrix is ​​in a high-noise region or a low-noise region.

[0116] Step 2: If the noise error value is greater than the noise error threshold, calculate the noise error adjustment intensity based on the noise error value.

[0117] In this embodiment, if the noise error value is greater than the noise error threshold, it means that the noise in the latest first channel feature matrix is ​​in a high-noise region. To ensure the denoising effect, the correction amplitude is reduced, i.e., the noise error adjustment intensity is calculated based on the noise error value. The noise error adjustment intensity is less than 1.

[0118] Here, the noise error adjustment intensity is calculated based on the noise error value, including:

[0119] Substituting the noise error value into the following formula yields the noise error adjustment strength;

[0120] ;

[0121] in, Adjust the intensity for noise error. This is the noise error value. This represents the remaining number of processing attempts.

[0122] Step 3: If the noise error value is less than or equal to the noise error threshold, then set the noise error adjustment intensity to 1.

[0123] In this embodiment, if the noise error value is less than or equal to the noise error threshold, it means that the noise in the latest first channel feature matrix is ​​in the low noise region. In order to ensure the denoising efficiency, the noise error adjustment intensity is set to 1 to ensure complete noise transmission.

[0124] Specifically, the second channel feature matrix is ​​corrected based on the noise error adjustment intensity and the noise error value to obtain the latest first channel feature matrix, including:

[0125] Substituting the noise error adjustment strength and the noise error value into the following formula, we obtain the latest first channel feature matrix;

[0126] ;

[0127] in, This is the latest first channel feature matrix. This is the second channel feature matrix. The remaining number of processing attempts. For noise reduction networks, These are the trainable parameters for the denoising network. This is the latest first channel feature matrix obtained when the remaining processing times are t+1.

[0128] Here, the remaining processing steps are used to select an appropriate diffusion model time step using signal-to-noise ratio (SNR) information to optimize channel estimation efficiency.

[0129] S104. Based on all the first channel feature matrices obtained at the end, determine the target channel state matrix corresponding to the pilot information.

[0130] In this embodiment of the application, all the first channel feature matrices obtained are subjected to 2D-DWT inverse transformation to convert the estimation results from the angular domain back to the spatial domain, thereby obtaining the target channel state matrix.

[0131] Here, compared to Gaussian random sampling in traditional diffusion models, this application employs a dynamic cold diffusion method for sampling (S102 to S104), progressively applying the learned noise prediction model to denoise the channel. This sampling strategy considers prediction error information when processing noise, exhibiting stronger tolerance to environmental noise and avoiding the traditional random resampling process. Adjustments are made to each state until a new channel state consistent with reality is reconstructed. This process employs a deterministic inverse process until the desired signal-to-noise ratio (SNR) level is achieved.

[0132] In the embodiments of this application, the unique inverse process (S102 to S104) and forward process (S201 to S204) are mainly relied upon to achieve accurate estimation of channel state information under the influence of Gaussian noise. Compared with traditional channel estimation methods, the diffusion model DM has the advantage of maintaining high estimation accuracy in noisy environments. By reducing the steps in the inverse process, the computational delay can be effectively reduced, thereby significantly improving the timeliness of channel estimation, which is particularly critical in rapidly changing wireless environments.

[0133] Specifically, the diffusion model, through its Markov chain model, treats channel estimation as a task of generating latent variables, approximating it at each step using a parameterized Gaussian transfer function. Compared to other generative models, the diffusion model (DM) exhibits better asymptoticity and robustness in progressively approximating the true channel model estimate. This approximation capability enables DM to provide consistent and efficient channel estimation under different signal-to-noise ratio (SNR) conditions, especially in large-scale MIMO systems with limited resources and computational requirements. In contrast, traditional machine learning channel estimation models perform poorly in SNR scenarios outside the training set.

[0134] The diffusion model is designed not only to consider signal reconstruction accuracy but also to incorporate a low-latency computation strategy, enabling it to outperform traditional methods in large-scale MIMO systems. By combining channel distribution learning in the sparse angular domain with subband learning, DM achieves an effective integration of resource consumption and computational efficiency, providing a practical and highly innovative channel estimation solution for large-scale MIMO systems.

[0135] In conclusion, the application of DM in MIMO channel estimation not only improves the recovery quality of CSI, but also optimizes system computation and resource efficiency. Its innovation and technical effects have a profound impact and application potential in the context of current B5G communication technology.

[0136] Based on the same inventive concept, this application also provides a channel estimation device for wireless communication corresponding to the channel estimation method for wireless communication. Since the principle of the device in this application is similar to the channel estimation method for wireless communication described above, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0137] Reference Figure 3 The diagram shown is a schematic of a channel estimation device for wireless communication provided in an embodiment of this application. The channel estimation device for wireless communication includes:

[0138] Extraction module 301 is used to extract the first channel feature matrix for each scale and each direction from the initial channel state matrix contained in the pilot information;

[0139] The input module 302 is used to input the latest first channel feature matrix into the denoising network for each first channel feature matrix to obtain a second channel feature matrix; the denoising network is trained based on the actual channel state matrix and the complete noise channel feature matrix corresponding to the actual channel state matrix obtained by using noise scheduling parameters; the noise scheduling parameters are used to control the noise addition ratio.

[0140] Correction module 303 is used to correct the second channel feature matrix according to the noise error value between the first channel feature matrix and the second channel feature matrix to obtain the latest first channel feature matrix; and continue to input the latest first channel feature matrix into the denoising network until a preset number of denoising times is reached;

[0141] The determination module 304 is used to determine the target channel state matrix corresponding to the pilot information based on all the first channel feature matrices obtained at the end.

[0142] like Figure 4 As shown in the embodiment of this application, an electronic device 400 includes a processor 401, a memory 402, and a bus. The memory 402 stores machine-readable instructions executable by the processor 401. When the electronic device is running, the processor 401 communicates with the memory 402 via the bus. The processor 401 executes the machine-readable instructions to perform the steps of the channel estimation method for wireless communication as described above.

[0143] Specifically, the memory 402 and processor 401 mentioned above can be general-purpose memory and processor, without any specific limitations. When the processor 401 runs the computer program stored in the memory 402, it can execute the channel estimation method for wireless communication mentioned above.

[0144] Corresponding to the channel estimation method for wireless communication described above, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the channel estimation method for wireless communication described above.

[0145] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some communication interfaces; the indirect coupling or communication connection of devices or modules can be electrical, mechanical, or other forms.

[0146] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0147] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0148] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the information processing methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0149] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A channel estimation method for wireless communication, characterized in that, The method includes: Extract the first channel feature matrix for each scale and each direction from the initial channel state matrix contained in the pilot information; For each first channel feature matrix, the latest first channel feature matrix is ​​input into the denoising network to obtain the second channel feature matrix; the denoising network is trained based on the actual channel state matrix and the complete noise channel feature matrix corresponding to the actual channel state matrix obtained using noise scheduling parameters; the noise scheduling parameters are used to control the noise addition ratio. Based on the noise error value between the first channel feature matrix and the second channel feature matrix, the second channel feature matrix is ​​corrected to obtain the latest first channel feature matrix; and the latest first channel feature matrix is ​​continued to be input into the denoising network until the preset number of denoising times is reached. Based on all the first channel feature matrices obtained at the end, the target channel state matrix corresponding to the pilot information is determined; The denoising network is trained through the following steps: Extracting actual channel feature matrices at each scale and direction from the actual channel state matrix; iteratively adding noise to the actual channel feature matrices according to the noise scheduling parameters to obtain a completely noisy channel feature matrix; denoising and correcting the completely noisy channel feature matrix using the denoising network to obtain the recovered channel state matrix corresponding to the completely noisy channel feature matrix; updating the parameters in the denoising network with the minimum difference between the recovered channel state matrix and the actual channel state matrix as the optimization objective; and continuing to denoise and correct the completely noisy channel feature matrix using the denoising network until a preset iteration stopping condition is met. The step of iteratively adding noise to the actual channel feature matrix according to the noise scheduling parameters to obtain the complete noise channel feature matrix includes: Input the noise scheduling parameters and the actual channel feature matrix into the following formula to obtain the complete noise channel feature matrix; ； ； in, This is the complete noise channel feature matrix. Let be the noise scheduling parameters for the i-th iteration. The number of times noise is added to the preset iteration. Let J be the noise scheduling parameters for the (j+1)th iteration. The standard complex Gaussian noise value following a standard normal distribution in the j-th iteration. Let I be the covariance matrix of the standard normal distribution. This is the actual channel feature matrix.

2. The channel estimation method for wireless communication according to claim 1, characterized in that, The step of correcting the second channel feature matrix based on the noise error value between the first channel feature matrix and the second channel feature matrix to obtain the latest first channel feature matrix includes: The noise error adjustment intensity is determined based on the numerical relationship between the noise error value and the noise error threshold. The second channel feature matrix is ​​corrected based on the noise error adjustment intensity and the noise error value to obtain the latest first channel feature matrix.

3. The channel estimation method for wireless communication according to claim 2, characterized in that, The step of determining the noise error adjustment intensity based on the numerical relationship between the noise error value and the noise error threshold includes: Determine whether the noise error value is greater than the noise error threshold; If the noise error value is greater than the noise error threshold, then the noise error adjustment intensity is calculated based on the noise error value; If the noise error value is less than or equal to the noise error threshold, then the noise error adjustment intensity is set to 1.

4. The channel estimation method for wireless communication according to claim 3, characterized in that, The step of calculating the noise error adjustment intensity based on the noise error value includes: Substituting the noise error value into the following formula yields the noise error adjustment intensity; ； in, Adjust the intensity for noise error. This is the noise error value. This represents the remaining number of processing attempts.

5. The channel estimation method for wireless communication according to claim 2, characterized in that, The step of correcting the second channel feature matrix based on the noise error adjustment intensity and the noise error value to obtain the latest first channel feature matrix includes: Substituting the noise error adjustment intensity and the noise error value into the following formula, the latest first channel feature matrix is ​​obtained; ； in, This is the latest first channel feature matrix. This is the second channel feature matrix. The remaining number of processing attempts. For noise reduction networks, These are the trainable parameters for the denoising network. This is the latest first channel feature matrix obtained when there are t+1 remaining processing iterations. Adjust the intensity for noise error. This represents the noise error value.

6. A channel estimation device for wireless communication, characterized in that, The device includes: The extraction module is used to extract the first channel feature matrix for each scale and each direction from the initial channel state matrix contained in the pilot information; The input module is used to input the latest first channel feature matrix into the denoising network for each first channel feature matrix to obtain a second channel feature matrix; the denoising network is trained based on the actual channel state matrix and the complete noise channel feature matrix corresponding to the actual channel state matrix obtained by using noise scheduling parameters; the noise scheduling parameters are used to control the noise addition ratio. The correction module is used to correct the second channel feature matrix according to the noise error value between the first channel feature matrix and the second channel feature matrix to obtain the latest first channel feature matrix; and continue to input the latest first channel feature matrix into the denoising network until the preset number of denoising times is reached; The determination module is used to determine the target channel state matrix corresponding to the pilot information based on all the first channel feature matrices obtained at the end; The denoising network is trained through the following steps: Extracting actual channel feature matrices at each scale and direction from the actual channel state matrix; iteratively adding noise to the actual channel feature matrices according to the noise scheduling parameters to obtain a completely noisy channel feature matrix; denoising and correcting the completely noisy channel feature matrix using the denoising network to obtain the recovered channel state matrix corresponding to the completely noisy channel feature matrix; updating the parameters in the denoising network with the minimum difference between the recovered channel state matrix and the actual channel state matrix as the optimization objective; and continuing to denoise and correct the completely noisy channel feature matrix using the denoising network until a preset iteration stopping condition is met. The step of iteratively adding noise to the actual channel feature matrix according to the noise scheduling parameters to obtain the complete noise channel feature matrix includes: Input the noise scheduling parameters and the actual channel feature matrix into the following formula to obtain the complete noise channel feature matrix; ； ； in, This is the complete noise channel feature matrix. Let be the noise scheduling parameters for the i-th iteration. The number of times noise is added to the preset iteration. Let J be the noise scheduling parameters for the (j+1)th iteration. The standard complex Gaussian noise value following a standard normal distribution in the j-th iteration. Let I be the covariance matrix of the standard normal distribution. This is the actual channel feature matrix.

7. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is in operation, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the channel estimation method for wireless communication as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the channel estimation method for wireless communication as described in any one of claims 1 to 5.

Images