Channel parameter estimation method, device, electronic device and storage medium

Abstract

The present application provides a channel parameter estimation method, apparatus, electronic device, and storage medium. The method utilizes a transmission signal, a target estimation channel parameter, a noise parameter, and a preset target smart reflector phase to perform processing through a sampling inference algorithm. The method obtains the target estimation channel parameter after t-1 rounds of processing. If the t-1 rounds of processing do not reach a preset processing threshold, the channel parameters after t-1 rounds of processing are used as the target estimation channel parameters. An iterative inference process is performed based on the transmission signal, the channel parameters after t-1 rounds of processing, and the target smart reflector phase, such that the smart reflector phase continuously approaches the true phase. Accurate inference can be performed based on this process. When the channel parameters after the latest processing round are obtained and the latest processing round reaches the preset processing threshold, the channel parameters after the latest processing round are used as the target channel parameters for the smart reflector, thereby ensuring the accuracy of channel parameter estimation.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a channel parameter estimation method, device, electronic device and storage medium. Background Art

[0002] Smart reflective surfaces can dynamically adjust the phase of reflective arrays based on environmental conditions and user needs. Large-scale reflective arrays can improve communication coverage, spectrum efficiency, and communication service quality. However, the planning of large-scale reflective arrays requires real-time channel parameter support. Therefore, channel parameter estimation is crucial for smart reflective surfaces.

[0003] However, the introduction of smart reflective surfaces results in more channel parameters needing to be estimated. When the number of pilot signals is limited, the difficulty of estimating channel parameters is greatly increased, resulting in a decrease in the accuracy of channel parameter estimation. Summary of the Invention

[0004] In view of this, the purpose of this application is to propose a channel parameter estimation method, device, electronic device and storage medium to solve the above technical problems.

[0005] Based on the above objectives, a first aspect of the present application provides a channel parameter estimation method, which is applied to a channel parameter estimation system, wherein the system includes a user terminal, a smart reflecting surface, and a base station connected in communication, and the method includes:

[0006] In response to the base station receiving a transmission signal, wherein the transmission signal is a signal formed by the smart reflector transmission based on an initial signal sent by the user terminal, obtaining a target channel parameter for estimation for t rounds of processing corresponding to the transmission signal;

[0007] Inputting the target estimation channel parameters into a pre-trained network model to obtain noise parameters;

[0008] The transmission signal, the target estimation channel parameter, the noise parameter, and the preset target smart reflector phase are processed by a sampling inference algorithm to obtain the channel parameter of the smart reflector in t-1 rounds of processing;

[0009] In response to the t-1 rounds of processing reaching a preset processing number threshold, the channel parameters of the t-1 rounds of processing are used as target channel parameters of the smart reflective surface; or

[0010] In response to the t-1 rounds of processing not reaching a preset processing number threshold, the channel parameters of the t-1 rounds of processing are used as target estimation channel parameters, and an iterative reasoning process is performed based on the transmission signal, the channel parameters of the t-1 rounds of processing, and the phase of the target smart reflector. The iterative reasoning process is as follows:

[0011] Processing the target smart reflector surface phase by a phase update algorithm based on the transmission signal, the channel parameters of the t-1 rounds of processing, and the target smart reflector surface phase to obtain the smart reflector surface phase of the t-1 rounds of processing;

[0012] Using the phase of the smart reflecting surface at the t-1th round of processing as the target smart reflecting surface phase, using the channel parameters at the t-1th round of processing as the target estimated channel parameters, and using the transmission signal, the channel parameters at the t-1th round of processing, the noise parameters, and the phase of the smart reflecting surface at the t-1th round of processing to obtain the channel parameters of the smart reflecting surface at the t-2th round of processing by performing processing through a sampling inference algorithm;

[0013] In response to the t-2 rounds of processing not reaching a preset processing number threshold, the iterative reasoning process is repeatedly performed until the channel parameters of the latest round of processing are obtained, and the latest round of processing reaches the preset processing number threshold, and the channel parameters of the latest round of processing are used as the target channel parameters of the smart reflecting surface.

[0014] Based on the same inventive concept, a second aspect of the present application provides a channel parameter estimation device, which is provided in a channel parameter estimation system, wherein the system includes a user terminal, a smart reflecting surface, and a base station in communication connection, and the device includes:

[0015] an acquisition module configured to, in response to the base station receiving a transmission signal, obtain a target channel parameter for estimation corresponding to t rounds of processing corresponding to the transmission signal, wherein the transmission signal is a signal formed by the smart reflection oriented transmission based on an initial signal sent by the user terminal;

[0016] An input module is configured to input the target estimation channel parameters into a pre-trained network model to obtain noise parameters;

[0017] The sampling and inference module is configured to use the transmission signal, the target estimation channel parameters, the noise parameters, and the preset target smart reflector phase to process through a sampling and inference algorithm to obtain the channel parameters of the smart reflector in t-1 rounds of processing; in response to the t-1 rounds of processing reaching a preset processing number threshold, use the channel parameters of the t-1 rounds of processing as the target channel parameters of the smart reflector; or, in response to the t-1 rounds of processing not reaching the preset processing number threshold, use the channel parameters of the t-1 rounds of processing as the target estimation channel parameters, and perform an iterative inference process based on the transmission signal, the channel parameters of the t-1 rounds of processing, and the target smart reflector phase, the iterative inference process being as follows:

[0018] The iterative reasoning module is configured to process the transmission signal, the channel parameters of the t-1 round of processing, and the target smart reflector phase through a phase updating algorithm to obtain the smart reflector phase of the t-1 round of processing; use the smart reflector phase of the t-1 round of processing as the target smart reflector phase, use the channel parameters of the t-1 round of processing as the target channel parameters for estimation, and use the transmission signal, the channel parameters of the t-1 round of processing, the noise parameters, and the smart reflector phase of the t-1 round of processing through a sampling inference algorithm to obtain the channel parameters of the smart reflector at the t-2 round of processing;

[0019] The repetition execution module is configured to, in response to the t-2 rounds of processing not reaching a preset processing number threshold, repeatedly execute the iterative reasoning process until the channel parameters of the latest round of processing are obtained, and the latest round of processing reaches the preset processing number threshold, and use the channel parameters of the latest round of processing as the target channel parameters of the smart reflecting surface.

[0020] Based on the same inventive concept, the third aspect of this application provides an electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable by the processor, wherein the processor implements the method described in the first aspect above when executing the computer program.

[0021] Based on the same inventive concept, the fourth aspect of the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the method described in the first aspect above.

[0022] As can be seen from the above, the channel parameter estimation method, apparatus, electronic device, and storage medium provided in the present application, when a base station receives a transmission signal sent by a user terminal through a smart reflector, obtains the target estimation channel parameters for t rounds of processing corresponding to the transmission signal, then inputs the target estimation channel parameters into a pre-trained network model to obtain noise parameters. The transmission signal, the target estimation channel parameters, the noise parameters, and the preset target smart reflector phase are then processed using a sampling inference algorithm to obtain the channel parameters of the smart reflector in t-1 rounds of processing. If the t-1 round of processing does not reach the preset processing number threshold, the channel parameters of the t-1 round of processing are used as the target estimation channel parameters. An iterative inference process is performed based on the transmission signal, the channel parameters of the t-1 round of processing, and the target smart reflector phase, so that the smart reflector phase continuously approaches the true phase. Therefore, accurate inference can be performed based on this. When the channel parameters for the latest round of processing are obtained and the latest round of processing reaches the preset processing number threshold, the channel parameters for the latest round of processing are used as the target channel parameters for the smart reflector, thereby ensuring the accuracy of the channel parameter estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in this application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are merely embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0024] Figure 1 This is a flow chart of a channel parameter estimation method according to an embodiment of the present application;

[0025] Figure 2A Schematic diagram of a communication system assisted by a smart reflective surface according to an embodiment of the present application;

[0026] Figure 2B A schematic diagram of a diffusion model according to an embodiment of the present application;

[0027] Figure 2C This is a schematic diagram of the initial network model architecture of an embodiment of the present application;

[0028] Figure 2D This is a schematic diagram of the initial network model training process of an embodiment of the present application;

[0029] Figure 2E This is a schematic diagram of the progressive distillation training process of an embodiment of the present application;

[0030] Figure 2F Schematic diagram of the channel parameter estimation process according to an embodiment of the present application;

[0031] Figure 3 This is a structural block diagram of a channel parameter estimation device according to an embodiment of the present application;

[0032] Figure 4 A schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the objectives, technical solutions and advantages of this application more clear, this application is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0034] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should have the usual meanings understood by people with ordinary skills in the field to which this application belongs. The "first", "second" and similar words used in the embodiments of the present application do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0035] It is understandable that before using the technical solutions of each embodiment of this application, the type, scope of use, usage scenarios, etc. of the personal information involved will be informed to the user in an appropriate manner, and the user's authorization will be obtained.

[0036] For example, in response to receiving a user's active request, a prompt message is sent to the user to clearly inform the user that the operation requested will require the acquisition and use of the user's personal information. Thus, the user can independently choose whether to provide personal information to the electronic device, application, server, storage medium, or other software or hardware that performs the operation of the technical solution of this application based on the prompt message.

[0037] As an optional but non-limiting implementation, in response to a user's active request, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. Furthermore, the pop-up window may also contain a selection control for the user to select "agree" or "disagree" to provide personal information to the electronic device.

[0038] It is understandable that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this application. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this application.

[0039] By planning large-scale reflector arrays, systems can improve coverage, spectral efficiency, and service quality. Smart reflector technology is hailed as a key technology for future 5G / 6G networks. Planning large-scale reflector arrays requires real-time channel parameter support, making channel estimation crucial for smart reflector surfaces. However, the introduction of smart reflector surfaces requires more channel parameters to be estimated, significantly increasing the difficulty of channel estimation given the limited number of system pilots.

[0040] Furthermore, the intelligent reflector can dynamically adjust the reflector array phase based on environmental conditions and user needs, thereby improving spectral efficiency, enhancing communication coverage, and enhancing the user experience. Phase adjustment of the intelligent reflector depends on channel parameters. In uplink mode, the base station can estimate the current channel parameters using pilot signals.

[0041] Current technologies for intelligent reflector cascade channel estimation fall into two main categories: those based on traditional statistical models and those based on deep learning. Deep learning methods can learn channel priors from complex environments, effectively addressing the mismatch between traditional statistical channel models and these environments. However, most current deep learning solutions directly learn the mapping from received signals to channel information, which results in low learning efficiency. For example, residual networks are used to learn the mapping from least squares estimation results to channel state information, thereby obtaining a denoised cascade channel. Alternatively, attention mechanisms are introduced into neural networks to further enhance channel feature extraction and improve the accuracy of channel state information. However, these methods are limited by an insufficient number of pilot signals and the influence of noise perturbations, which increases learning difficulty. Furthermore, iterative deep learning solutions have been proposed in related technologies, integrating trained network models into an iterative architecture. However, these methods fail to maintain strict consistency between each iterative inference step during training, resulting in limited learning performance. For example, integrating neural network models into iterative threshold algorithms gradually updates channel state information, thereby improving estimation accuracy. However, this solution does not maintain consistency with each iterative process when training the neural network. Instead, it selects noisy channel samples for unified training, resulting in a decrease in the efficiency of the iterative process. In addition, there are currently channel estimation solutions based on diffusion models that can keep the training process consistent with the iterative process. For example, a corresponding forward process and reverse inference process are constructed, where the forward process provides training data for the reverse process, ensuring that the deep learning model and the reverse iterative process are strictly consistent. Alternatively, the diffusion model can be applied to underwater acoustic communication channel estimation scenarios. However, the above solution requires too many iterations and the required computational complexity is too high, making it difficult to meet the needs of obtaining channel state information in real time. At the same time, smart reflectors are usually made of inexpensive materials, so the reflector surface in the actual hardware system will have phase noise. The above solution does not consider processing phase noise, which affects the robustness of the solution.

[0042] In summary, deep learning-based solutions in related technologies generally directly learn the mapping from received signals to channel state information, which leads to reduced learning efficiency. Iterative deep learning solutions have problems such as the training process and the iteration process not being strictly consistent. Channel estimation research solutions based on diffusion models require a large number of iterations, resulting in very high computational complexity, which makes them difficult to deploy in real-time systems. At the same time, because smart reflectors are usually made of inexpensive materials, their phase is not very accurate in actual hardware systems and is subject to phase noise. However, current channel estimation research solutions based on diffusion models do not consider the phase noise of smart reflectors, resulting in reduced channel estimation accuracy.

[0043] An embodiment of the present application provides a channel parameter estimation method, which uses a transmission signal, a channel parameter for target estimation, a noise parameter, and a preset target smart reflector phase to be processed through a sampling inference algorithm to obtain the channel parameters of the smart reflector in t-1 rounds of processing. If the number of t-1 rounds of processing does not reach a preset processing number threshold, the channel parameters of the t-1 rounds of processing are used as the channel parameters for target estimation. An iterative inference process is performed based on the transmission signal, the channel parameters of the t-1 rounds of processing, and the target smart reflector phase, so that the smart reflector phase continuously approaches the true phase. Therefore, accurate inference can be performed based on this. When the channel parameters of the latest round of processing are obtained and the latest round of processing reaches the preset processing number threshold, the channel parameters of the latest round of processing are used as the target channel parameters of the smart reflector, thereby ensuring the accuracy of channel parameter estimation.

[0044] like Figure 1 As shown, the method of this embodiment is applied to a channel parameter estimation system, wherein the system includes a user terminal, a smart reflecting surface, and a base station in communication connection. The method of this embodiment includes:

[0045] Step 101: In response to the base station receiving a transmission signal, wherein the transmission signal is a signal formed by the intelligent reflection-oriented transmission based on the initial signal sent by the user terminal, a target estimation channel parameter for t rounds of processing corresponding to the transmission signal is obtained.

[0046] In this step, if Figure 2A As shown in FIG, the base station is equipped with N antennas, the user terminal is equipped with a single antenna, the smart reflector array consists of M passive array elements, and both the base station and the smart reflector use a planar array with a half-wavelength interval.

[0047] This application uses the wireless channel model (Saleh-Valenzuela) to model the smart reflector-base station channel G and the user terminal-smart reflector channel h:

[0048]

[0049]

[0050] Among them, L G 、L h is the number of multipaths, Respectively represent the l1th path sub-channel of G and the l2th path sub-channel gain amplitude of h, φ, Represent the azimuth and elevation angles respectively. represents the base station side antenna array steering vector, where represents the Kronecker product. is defined as:

[0051]

[0052]

[0053] Among them, N1, N2 (N = N1 × N2) represent the number of antennas in the horizontal and vertical directions. In the above formula and It is constructed in the same way.

[0054] Given the phase θ of the smart reflector k =[θ 1，k ,…,θ M，k ] T and pilot signal S k , the transmission signal Y received by the base station k It can be expressed as

[0055] in, is complex Gaussian noise, δ 2 is the variance, the cascade channel can be expressed as H cas =Gdiag(h). After collecting the Q-slot pilot signal, the transmission signal received by the base station is Y=H cas Θ+W. Among them, Since smart reflectors are usually made of cheap materials, they introduce phase noise, and the actual phase With the preset phase θ i，j The relationship is: Represents phase noise. Therefore, the purpose of channel estimation for the intelligent reflector cascade is to estimate the transmission signal Y received by the base station and the preset phase θ i，j , estimate the cascade channel (i.e., target channel parameters).

[0056] Step 102: Input the target estimation channel parameters into a pre-trained network model to obtain noise parameters.

[0057] In this step, the present application transforms the intelligent reflector cascade channel estimation problem into a continuous sampling process. By establishing a forward process and a reverse process, the training and sampling processes are strictly consistent, greatly improving the estimation accuracy of the network model.

[0058] Therefore, inputting the target estimated channel parameters into a pre-trained network model can make the obtained noise parameters more accurate.

[0059] Step 103 : Process the transmission signal, the target estimation channel parameters, the noise parameters, and the preset target smart reflector phase through a sampling inference algorithm to obtain the channel parameters of the smart reflector in t-1 rounds of processing.

[0060] In this step, the application will condition the gradient Integrating into the sampling process, the conditional gradient is calculated as follows:

[0061]

[0062] The sampling inference algorithm based on conditional gradient is as follows:

[0063]

[0064] Among them, λ1 is an empirical parameter, is the result estimated by the network,

[0065] By processing the transmission signal, target estimation channel parameters, noise parameters and the preset target smart reflector phase through the sampling inference algorithm, more accurate channel parameters of the smart reflector in t-1 rounds of processing can be obtained.

[0066] Step 104: In response to the t-1 round of processing times reaching a preset processing times threshold, the channel parameters of the t-1 round of processing times are used as the target channel parameters of the smart reflective surface. Or,

[0067] In this step, the preset processing times threshold can be set according to the specific situation and is not specifically limited here. For example, the preset processing times threshold is 1. When the number of t-1 rounds of processing reaches 1, the channel parameters of the t-1 rounds of processing are used as the target channel parameters of the smart reflector surface.

[0068] Step 105: In response to the t-1 round of processing not reaching a preset processing number threshold, the channel parameters of the t-1 round of processing are used as target estimation channel parameters, and an iterative reasoning process is performed based on the transmission signal, the channel parameters of the t-1 round of processing, and the phase of the target smart reflector. The iterative reasoning process is as follows:

[0069] Step 1051 : Process the transmission signal, the channel parameters of the t-1 rounds of processing, and the target smart reflector phase using a phase update algorithm to obtain the smart reflector phase of the t-1 rounds of processing.

[0070] In this step, when the number of processing rounds t-1 does not reach the preset processing number threshold, the presence of phase noise causes the above-mentioned sampling process inference algorithm to be inaccurate, seriously affecting the sampling performance. Therefore, this application also incorporates the solution process of the true phase into the sampling process. Similar to the conditional gradient solution, the phase is updated based on the transmission signal, the channel parameters of the t-1 processing rounds, and the target smart reflector phase through the phase update algorithm. This allows the smart reflector phase to continuously approach the true phase, and the smart reflector phase of the t-1 processing rounds is obtained.

[0071] Step 1052: Use the phase of the smart reflector after t-1 rounds of processing as the target smart reflector phase, use the channel parameters after t-1 rounds of processing as the target estimated channel parameters, and use the transmission signal, the channel parameters after t-1 rounds of processing, the noise parameters, and the phase of the smart reflector after t-1 rounds of processing to obtain the channel parameters of the smart reflector after t-2 rounds of processing.

[0072] In this step, after the phase update, the transmission signal, the channel parameters of the t-1th round of processing, the noise parameters, and the phase of the smart reflector of the t-1th round of processing are processed through a sampling inference algorithm to obtain more accurate channel parameters of the smart reflector in the t-2th round of processing.

[0073] Step 106: In response to the t-2 rounds of processing not reaching the preset processing number threshold, the iterative reasoning process is repeated until the channel parameters of the latest round of processing are obtained, and the latest round of processing reaches the preset processing number threshold, and the channel parameters of the latest round of processing are used as the target channel parameters of the smart reflective surface.

[0074] In this step, if the number of t-2 rounds of processing does not reach the preset processing number threshold, the above iterative reasoning process is repeated until the number of the latest round of processing reaches the preset processing number threshold. For example, if the number of the latest round of processing reaches 1, the channel parameters of the latest round of processing are used as the target channel parameters of the smart reflection surface to ensure the accuracy of the channel parameter estimation.

[0075] According to the above scheme, when a base station receives a transmission signal sent by a user terminal via a smart reflector, it obtains target channel parameters for estimation corresponding to t rounds of processing of the transmission signal. The target channel parameters for estimation are then input into a pre-trained network model to obtain noise parameters. A sampling inference algorithm is then used to process the transmission signal, the target channel parameters for estimation, the noise parameters, and a preset target smart reflector phase to obtain the channel parameters for the smart reflector at t-1 rounds of processing. If the number of processing rounds t-1 does not reach a preset processing number threshold, the channel parameters for the t-1 round of processing are used as the target channel parameters for estimation. An iterative inference process is then performed based on the transmission signal, the channel parameters for the t-1 round of processing, and the target smart reflector phase, so that the smart reflector phase continuously approaches the true phase. Therefore, accurate inference can be performed based on this process. When the channel parameters for the latest processing round are obtained and the latest processing round reaches the preset processing number threshold, the channel parameters for the latest processing round are used as the target channel parameters for the smart reflector, thereby ensuring the accuracy of channel parameter estimation.

[0076] In some embodiments, step 103 includes:

[0077] The channel parameter of the smart reflecting surface in the t-1th round of processing is determined by the following formula according to the transmission signal, the target estimation channel parameter, the noise parameter and the preset target smart reflecting surface phase:

[0078]

[0079] Among them, x t-1 represents the channel parameter of the smart reflective surface in the t-1th round of processing, α t =1-β t , β t represents the variance, x t represents the channel parameters used for target estimation, represents the noise parameter, λ1 represents the preset first empirical parameter, Θ represents the phase of the target intelligent reflector, Y represents the transmission signal, z obeys the standard normal distribution, and H represents the conjugate transpose.

[0080] In the above scheme, this application will condition gradient Integrating into the sampling process, the conditional gradient is calculated as follows:

[0081]

[0082] The sampling inference algorithm based on conditional gradient is as follows:

[0083]

[0084] By processing the transmission signal, target estimation channel parameters, noise parameters and the preset target smart reflector phase through the sampling inference algorithm, more accurate channel parameters of the smart reflector in t-1 rounds of processing can be obtained.

[0085] In some embodiments, step 1051 includes:

[0086] The phase of the smart reflector for the t-1 round of processing is determined by the following formula based on the transmission signal, the channel parameter for the t-1 round of processing, and the target smart reflector phase:

[0087] Θ t-1 =λ2x t H (Yx t Θ t )+Θ t .

[0088] Among them, Θ t-1 represents the phase of the intelligent reflector after t-1 rounds of processing, λ2 represents the preset second empirical parameter, and x t represents the channel parameters used for target estimation, H represents the conjugate transpose, Y represents the transmitted signal, Θ t Indicates the phase of the target intelligent reflector.

[0089] In the above scheme, when the number of processing rounds t-1 does not reach the preset processing number threshold, the presence of phase noise causes the above sampling process inference algorithm to be inaccurate, seriously affecting the sampling performance. Therefore, this application also integrates the solution process of the true phase into the sampling process. Similar to the conditional gradient solution, the phase is updated based on the transmission signal, the channel parameters of the t-1 processing rounds, and the target smart reflector phase through the phase update algorithm. This allows the smart reflector phase to continuously approach the true phase, and a more accurate smart reflector phase for the t-1 processing rounds is obtained.

[0090] In some embodiments, step 1052 includes:

[0091] The channel parameter of the smart reflecting surface at the t-2 round of processing is determined by the following formula using the transmission signal, the channel parameter at the t-1 round of processing, the noise parameter, and the phase of the smart reflecting surface at the t-1 round of processing:

[0092]

[0093] Among them, x t-2 represents the channel parameter of the smart reflective surface in t-2 rounds of processing, α t =1-β t , βt represents the variance, x t-1 represents the channel parameter of the smart reflective surface in the t-1th round of processing, represents the noise parameter, λ1 represents the preset first empirical parameter, Θ t-1 represents the phase of the smart reflector after t-1 rounds of processing, Y represents the transmitted signal, Z obeys the standard normal distribution, and H represents the conjugate transpose.

[0094] In the above scheme, after the phase update, the transmission signal, the channel parameters of the t-1 round of processing, the noise parameters and the phase of the smart reflector of the t-1 round of processing are processed through the sampling inference algorithm to obtain more accurate channel parameters of the smart reflector in the t-2 round of processing.

[0095] In some embodiments, in step 102, the training process of the network model includes:

[0096] Step A1: construct a diffusion model and an initial network model, determine noise parameters of the diffusion model in the forward process, and generate noisy channel parameters based on the noise parameters.

[0097] Step A2: determining a first loss function based on the noise parameter and the noisy channel parameter.

[0098] Step A3: input the noise channel parameters into the initial network model, and train and adjust the diffusion model based on the minimization result obtained in the process of minimizing the first loss function to obtain a trained initial network model.

[0099] Step A4: Perform progressive distillation training on the trained initial network model to obtain the network model.

[0100] In the above scheme, if Figure 2B As shown, this application transforms the smart reflector cascade channel estimation problem into a continuous sampling process. By establishing a forward process and a conditional reverse process, the training and sampling processes are strictly consistent, greatly improving the estimation accuracy.

[0101] Specifically, the forward process is to gradually add noise to x0, and finally obtain pure Gaussian noise x T (i.e., noise channel parameters), where T is the number of transfer steps (i.e., number of processing times). Each transfer step follows a Markov process:

[0102]

[0103] Among them, β t represents the variance. After a sufficient number of steps, x TIt will eventually obey the standard normal distribution. Then p(x t |x0) can be derived as

[0104] in, α t =1-β t The forward process generates training data (i.e., noisy channel parameters) for subsequent network model (i.e., initial network model) training.

[0105] The conditional reverse process is described as follows. The conditional reverse process is symmetrical with the forward process. T Gradually restore x0. The reverse process transition probability is:

[0106]

[0107] Among them, for the above distribution, the model based on deep learning is an efficient way to determine the distribution mean and covariance parameters, which can avoid the large amount of computational overhead and corresponding computational delay required for sampling inference in traditional methods. The initial network model of this application uses a U-shaped network to learn the above parameters, and its learning goal is to maximize the likelihood function p η (x0) is derived. p η (x0) is estimated by the U-shaped network. The final training loss function of the U-shaped network (i.e., the first loss function) is shown as follows:

[0108]

[0109] in, f η (.) represents a U-shaped network. The U-shaped network structure is as follows Figure 2C As shown, this application will x t The real and imaginary parts of the noise channel parameters are separated and stacked as the input of the network. The architecture is divided into 4 layers, where the resolution of the feature map is halved (from N×2M to (N / 8)x(2M / 8)). In the left half of the U-shaped network, x t A 2×2 max pooling operation halves the feature map resolution, followed by convolution, activation functions, and other processing to generate channel features. The right half of the U-shaped network is symmetrical: the feature map on the right side of each layer has the same resolution as the left side. By concatenating the feature maps on both sides of the same layer, high-level and low-level features are integrated, improving feature extraction.

[0110] The U-shaped network training process is as follows Figure 2D As shown:

[0111] Step 201: Initialize x0=H cas (i.e. the initial channel parameters of the diffusion model in the forward process), set the number of training termination rounds, {αt}, {β t}.

[0112] Step 202: Record the current training round number and randomly sample t to {1, ..., T}.

[0113] Step 203: Generate noise interference x t The noise ε (i.e., noise parameter).

[0114] Step 204: Based on the training loss function (ie, the first loss function) and the data in step 203, the network model (ie, the initial network model) is trained, and the number of training rounds is increased by one.

[0115] Step 205: Repeat steps 202 to 204 until the model training converges or the maximum number of rounds is met (i.e., the stopping condition), and then execute step 206: output the trained network model (i.e., the trained initial network model). Alternatively, if the model training does not converge or the maximum number of rounds is not met (i.e., the stopping condition), then return to step 202.

[0116] Then, the trained initial network model is subjected to progressive distillation training to improve sampling efficiency, shorten the number of sampling steps, and obtain the network model.

[0117] In some embodiments, in step A1, determining the noise channel parameters of the diffusion model includes:

[0118] Step A11: Acquire initial channel parameters of the diffusion model in the forward process.

[0119] Step A12: performing noise processing on the initial channel parameters according to a Markov process to obtain the noisy channel parameters of the diffusion model.

[0120] In the above scheme, if Figure 2B As shown, this application transforms the smart reflector cascade channel estimation problem into a continuous sampling process. By establishing a forward process and a conditional reverse process, the training and sampling processes are strictly consistent, greatly improving the estimation accuracy.

[0121] Specifically, the forward process is to gradually add noise to x0 (i.e., the initial channel parameters), and finally obtain pure Gaussian noise x T (i.e., noise channel parameters), where T is the number of transfer steps (i.e., number of processing times). Each transfer step follows a Markov process:

[0122]

[0123] In some embodiments, step A4 includes:

[0124] Step B1: Obtain training channel parameters and noise parameters output by the trained initial network model in t-1 rounds of processing, and construct a student model corresponding to the trained initial network model.

[0125] Step B2: determining a second loss function based on the training channel parameters and the noise parameters output by the trained initial network model in t-1 rounds of processing.

[0126] Step B3: input the training channel parameters into the student model, and perform training and adjustment on the student model based on the minimization result obtained in the process of minimizing the second loss function to obtain a network model.

[0127] In this approach, we design a progressive distillation framework, using the trained initial noise model as a teacher model to guide the training of its student model, thereby continuously halving the number of sampling steps. The number of sampling steps decreases exponentially with the number of distillation rounds. Therefore, the progressive distillation architecture effectively reduces computational complexity while maintaining performance, making it suitable for deployment in real-time systems.

[0128] This application further improves the trained initial network model, improves sampling efficiency, and shortens the number of sampling steps. The student model of the trained initial network model is gradually trained through a progressive distillation architecture. In the first round of distillation, the trained initial network model is regarded as a teacher model to guide the training of the student model. The student model uses the same network architecture as the teacher model and is initialized. At time step t, the input of the student model is:

[0129]

[0130] The learning goal of the student model is the two-step sampling output of the teacher model, namely:

[0131]

[0132]

[0133] The student model is required to output the same mean Therefore, the learning objective of the student model (i.e., the second loss function) is summarized and derived as follows:

[0134]

[0135] After the first round of distillation, the trained student model is regarded as the teacher model in the next round of distillation to guide the training of the new student model. Each round of distillation will halve the number of sampling steps.

[0136] Progressive distillation process Figure 2E As shown:

[0137] Step 301: Initialize the student model and assign it the weights of the existing teacher model.

[0138] Step 302: Train the student model. The learning objective is as shown in the above formula. Train the student model.

[0139] Step 303: Obtain the trained student model and regard it as the teacher model in the next round of distillation process.

[0140] Step 304: Repeat steps 301 to 303 until the student model performance shows a significant decline, i.e., the stopping condition is met, then execute step 305: output the diffusion model (i.e., network model) after progressive distillation. Alternatively, if the student model performance does not show a significant decline, i.e., the stopping condition is not met, then return to step 301.

[0141] In some embodiments, unlike the image generation task, channel parameter estimation requires estimating channel parameters based on the transmission signal received by the base station and the preset target intelligent reflector phase, wherein the channel parameters are used to represent the channel state information. Therefore, this application will conditional gradient conditional gradient Integrating into the sampling process, the conditional gradient is calculated as follows:

[0142]

[0143] The sampling inference algorithm based on conditional gradient is as follows:

[0144]

[0145] Among them, λ1 is an empirical parameter, is the result estimated by the network,

[0146] However, due to the presence of phase noise, the above sampling process is inaccurate, which seriously affects the sampling performance. Therefore, this application also integrates the solution process of the true phase into the sampling process, similar to the conditional gradient solution, and the phase update is as follows:

[0147] Θ t-1 =λ2x t H (Yx t Θ t )+Θ t

[0148] Therefore, the entire conditional sampling process is as follows Figure 2F As shown:

[0149] Step 401: Initialize parameters during sampling Set the time step number T, {αt}, {β t}.

[0150] Step 402: Record the time step t and calculate x according to the sampling inference algorithm t-1 .

[0151] Step 403: The smart reflector phase solution is updated, with the time step number t=t-1.

[0152] Step 404: Repeat steps 402 and 403 until t=1, then the stop condition is met, and step 405 is executed: output the estimated channel state information (i.e., channel parameters) Alternatively, if t is not equal to 1 and the stop condition is not satisfied, the process returns to and repeats step 402 .

[0153] It should be noted that the method of the embodiment of the present application can be performed by a single device, such as a computer or server. The method of this embodiment can also be applied in a distributed scenario and performed by multiple devices working together. In such a distributed scenario, one of the multiple devices may only perform one or more steps of the method of the embodiment of the present application, and the multiple devices will interact with each other to complete the method.

[0154] It should be noted that the above description is limited to some embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in an order different from that described in the above embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0155] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a channel parameter estimation device.

[0156] refer to Figure 3 The channel parameter estimation device is provided in a channel parameter estimation system, the system including a user terminal, an intelligent reflecting surface and a base station connected in communication, and the device includes:

[0157] An acquisition module 301 is configured to, in response to the base station receiving a transmission signal, wherein the transmission signal is a signal formed by the smart reflector-oriented transmission based on an initial signal sent by the user terminal, obtain a target channel parameter for t rounds of processing corresponding to the transmission signal;

[0158] An input module 302 is configured to input the target estimation channel parameters into a pre-trained network model to obtain noise parameters;

[0159] The sampling and inference module 303 is configured to use the transmission signal, the target estimation channel parameters, the noise parameters, and the preset target smart reflector phase to process through a sampling and inference algorithm to obtain the channel parameters of the smart reflector after t-1 rounds of processing; in response to the t-1 rounds of processing reaching a preset processing number threshold, use the channel parameters of the t-1 rounds of processing as the target channel parameters of the smart reflector; or, in response to the t-1 rounds of processing not reaching the preset processing number threshold, use the channel parameters of the t-1 rounds of processing as the target estimation channel parameters, and perform an iterative inference process based on the transmission signal, the channel parameters of the t-1 rounds of processing, and the target smart reflector phase. The iterative inference process is as follows:

[0160] The iterative reasoning module 304 is configured to process the transmission signal, the channel parameters of the t-1 processing round, and the target smart reflector phase using a phase updating algorithm to obtain the smart reflector phase of the t-1 processing round; use the smart reflector phase of the t-1 processing round as the target smart reflector phase, use the channel parameters of the t-1 processing round as the target estimation channel parameters, and use the transmission signal, the channel parameters of the t-1 processing round, the noise parameters, and the smart reflector phase of the t-1 processing round using a sampling inference algorithm to obtain the channel parameters of the smart reflector at the t-2 processing round.

[0161] The repetitive execution module 305 is configured to, in response to the t-2 rounds of processing not reaching a preset processing number threshold, repeatedly execute the iterative reasoning process until the channel parameters of the latest round of processing are obtained, and the latest round of processing reaches the preset processing number threshold, and use the channel parameters of the latest round of processing as the target channel parameters of the smart reflecting surface.

[0162] In some embodiments, the sampling inference module 303 is specifically configured to:

[0163] The channel parameter of the smart reflecting surface in the t-1th round of processing is determined by the following formula according to the transmission signal, the target estimation channel parameter, the noise parameter and the preset target smart reflecting surface phase:

[0164]

[0165] Among them, x t-1 represents the channel parameter of the smart reflective surface in the t-1th round of processing, α t =1-β t , β t represents the variance, x t represents the channel parameters used for target estimation, represents the noise parameter, λ1 represents the preset first empirical parameter, Θ represents the phase of the target intelligent reflector, Y represents the transmission signal, z obeys the standard normal distribution, and H represents the conjugate transpose.

[0166] In some embodiments, the iterative reasoning module 304 is specifically configured to:

[0167] The phase of the smart reflector for the t-1 round of processing is determined by the following formula based on the transmission signal, the channel parameter for the t-1 round of processing, and the target smart reflector phase:

[0168] Θ t-1 =λ2x t H (Yx t Θ t )+Θ t

[0169] Among them, Θ t-1 represents the phase of the intelligent reflector after t-1 rounds of processing, λ2 represents the preset second empirical parameter, and x t represents the channel parameters used for target estimation, H represents the conjugate transpose, Y represents the transmitted signal, Θ t Indicates the phase of the target intelligent reflector.

[0170] In some embodiments, the iterative reasoning module 304 is specifically configured to:

[0171] The channel parameter of the smart reflecting surface at the t-2 round of processing is determined by the following formula using the transmission signal, the channel parameter at the t-1 round of processing, the noise parameter, and the phase of the smart reflecting surface at the t-1 round of processing:

[0172]

[0173] Among them, x t-2 represents the channel parameter of the smart reflective surface in t-2 rounds of processing, α t =1-β t , β t represents the variance, x t-1 represents the channel parameter of the smart reflective surface in the t-1th round of processing, represents the noise parameter, λ1 represents the preset first empirical parameter, Θ t-1represents the phase of the smart reflector after t-1 rounds of processing, Y represents the transmitted signal, z obeys the standard normal distribution, and H represents the conjugate transpose.

[0174] In some embodiments, the channel parameter estimation apparatus further includes a training module, wherein the training module includes:

[0175] a noise-added channel parameter determination unit configured to construct a diffusion model and an initial network model, determine noise parameters of the diffusion model in a forward process, and generate noise-added channel parameters based on the noise parameters;

[0176] a loss function determining unit, configured to determine a first loss function based on the noise parameter and the noisy channel parameter;

[0177] a training adjustment unit configured to input the noise channel parameters into the initial network model, and perform training adjustment on the diffusion model based on a minimization result obtained during the minimization of the first loss function to obtain a trained initial network model;

[0178] The progressive distillation training unit is configured to perform progressive distillation training on the trained initial network model to obtain the network model.

[0179] In some embodiments, the noise channel parameter determination unit is specifically configured to:

[0180] Obtaining initial channel parameters of the diffusion model in the forward process;

[0181] The initial channel parameters are subjected to noise addition processing according to a Markov process to obtain the noisy channel parameters of the diffusion model.

[0182] In some embodiments, the progressive distillation training unit is specifically configured to:

[0183] Obtaining training channel parameters and noise parameters output by the trained initial network model in t-1 rounds of processing, and constructing a student model corresponding to the trained initial network model;

[0184] Determining a second loss function based on the training channel parameters and the noise parameters output by the trained initial network model in t-1 rounds of processing;

[0185] The training channel parameters are input into the student model, and the student model is trained and adjusted based on the minimization result obtained in the process of minimizing the second loss function to obtain a network model.

[0186] For the convenience of description, the above devices are described as being divided into various modules according to their functions. Of course, when implementing this application, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0187] The apparatus of the above embodiment is used to implement the corresponding channel parameter estimation method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be described in detail here.

[0188] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein when the processor executes the program, the channel parameter estimation method described in any of the above embodiments is implemented.

[0189] Figure 4 A more specific hardware structure diagram of an electronic device provided in this embodiment is shown. The device may include: a processor 401, a memory 402, an input / output interface 403, a communication interface 404, and a bus 405. The processor 401, the memory 402, the input / output interface 403, and the communication interface 404 are communicatively connected to each other within the device via the bus 405.

[0190] The processor 401 can be implemented using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0191] The memory 402 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage devices, dynamic storage devices, etc. The memory 402 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 402 and called and executed by the processor 401.

[0192] The input / output interface 403 is used to connect to input / output modules to implement information input and output. The input / output modules can be configured as components within the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. Input devices may include a keyboard, mouse, touch screen, microphone, various sensors, etc., and output devices may include a display, speaker, vibrator, indicator light, etc.

[0193] The communication interface 404 is used to connect to a communication module (not shown) to enable communication between the device and other devices. The communication module can communicate via a wired method (such as USB, network cable, etc.) or a wireless method (such as mobile network, WIFI, Bluetooth, etc.).

[0194] The bus 405 comprises a pathway for transmitting information between various components of the device (eg, the processor 401 , the memory 402 , the input / output interface 403 , and the communication interface 404 ).

[0195] It should be noted that although the above device only shows the processor 401, the memory 402, the input / output interface 403, the communication interface 404, and the bus 405, in a specific implementation, the device may also include other components necessary for normal operation. In addition, it will be understood by those skilled in the art that the above device may only include the components necessary to implement the embodiments of this specification, and does not necessarily include all the components shown in the figure.

[0196] The electronic device of the above embodiment is used to implement the corresponding channel parameter estimation method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be described in detail here.

[0197] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the channel parameter estimation method described in any of the above embodiments.

[0198] The computer-readable media of this embodiment include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0199] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the channel parameter estimation method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0200] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application (including the claims) is limited to these examples. Within the scope of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0201] In addition, for simplicity of description and discussion, and in order not to make the embodiment of the application difficult to understand, the known power supply / ground connection with integrated circuit (IC) chip and other components may or may not be shown in the accompanying drawings provided. In addition, the device can be shown in the form of a block diagram to avoid making the embodiment of the application difficult to understand, and this also takes into account the following fact, that is, the details of the embodiment of these block diagram devices are highly dependent on the platform to be implemented in the embodiment of the application (that is, these details should be fully within the scope of understanding of those skilled in the art). When specific details (for example, circuit) are set forth to describe exemplary embodiments of the application, it will be apparent to those skilled in the art that the embodiment of the application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.

[0202] Although the present invention has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may utilize the embodiments discussed.

[0203] The embodiments of the present application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the scope of protection of this application.

Claims

1. A channel parameter estimation method, characterized in that: Applied to a channel parameter estimation system, the system includes a user terminal, a smart reflecting surface, and a base station connected by a communication link, and the method includes: In response to the base station receiving a transmission signal, wherein the transmission signal is a signal formed by the intelligent reflection-oriented transmission based on the initial signal sent by the user terminal, obtaining a corresponding Channel parameters used to estimate the target number of rounds of processing; Inputting the target estimation channel parameters into a pre-trained network model to obtain noise parameters; The transmission signal, the target estimation channel parameter, the noise parameter and the preset target smart reflector phase are processed by a sampling inference algorithm to obtain the phase of the smart reflector. Channel parameters for the number of rounds of processing; In response to the When the number of rounds of processing reaches the preset processing number threshold, the The channel parameters of the number of rounds of processing are used as target channel parameters of the smart reflective surface; In response to the If the number of rounds of processing does not reach the preset processing number threshold, the The channel parameters of the round processing times are used as the target estimation channel parameters, based on the transmission signal, the The channel parameters of the number of rounds of processing and the phase of the target intelligent reflector are subjected to an iterative reasoning process, and the iterative reasoning process is as follows: Based on the transmission signal, the The channel parameters of the round processing times and the phase of the target intelligent reflector are processed by the phase update algorithm to obtain Smart reflector phase with round processing times; The The phase of the smart reflector with the highest number of rounds of processing is taken as the target phase of the smart reflector. The channel parameters of the number of rounds of processing are used as target estimation channel parameters, and the transmission signal, the The channel parameters of the number of rounds of processing, the noise parameters and the The phase of the smart reflector with the number of rounds of processing is processed by the sampling inference algorithm to obtain the phase of the smart reflector at Channel parameters for the number of rounds of processing; In response to the If the number of rounds of processing does not reach a preset processing number threshold, the iterative reasoning process is repeated until the channel parameters of the latest round of processing are obtained, and the latest round of processing reaches the preset processing number threshold, and the channel parameters of the latest round of processing are used as the target channel parameters of the smart reflecting surface; The transmission signal, the target estimation channel parameter, the noise parameter and the preset target smart reflector phase are processed by a sampling inference algorithm to obtain the smart reflector phase. Channel parameters for round processing times, including: The phase of the smart reflector surface is determined by the following formula according to the transmission signal, the target estimation channel parameter, the noise parameter and the preset target smart reflector surface phase. Channel parameters for round processing times: in, Indicates that the smart reflective surface is Channel parameters for the number of rounds of processing, , , , represents the variance, , represents the channel parameters used for target estimation, represents the noise parameter, represents the preset first empirical parameter, Indicates the phase of the target intelligent reflector surface, Indicates the transmission signal, Obeying the standard normal distribution, represents the conjugate transpose; The transmission signal, the The channel parameters of the round processing times and the phase of the target intelligent reflector are processed by the phase update algorithm to obtain Smart reflector phase with multiple rounds of processing, including: Based on the transmission signal, the The channel parameters of the round processing times and the phase of the target intelligent reflector are determined by the following formula Smart reflector phase with round processing times: in, express Smart reflector phase with round processing times, Represents the preset second empirical parameter, represents the channel parameters used for target estimation, represents the conjugate transpose, Indicates the transmission signal, Indicates the phase of the target intelligent reflector surface; said utilizing said transmission signal, said The channel parameters of the number of rounds of processing, the noise parameters and the The phase of the smart reflector with the number of rounds of processing is processed by the sampling inference algorithm to obtain the phase of the smart reflector at Channel parameters for round processing times, including: Using the transmission signal, the The channel parameters of the number of rounds of processing, the noise parameters and the The phase of the smart reflector after the number of rounds of processing is determined by the following formula: Channel parameters for round processing times: in, Indicates that the smart reflective surface is Channel parameters for the number of rounds of processing, , , , represents the variance, , Indicates that the smart reflective surface is Channel parameters for the number of rounds of processing, represents the noise parameter, represents the preset first empirical parameter, express Smart reflector phase with round processing times, Indicates the transmission signal, Obeying the standard normal distribution, represents the conjugate transpose.

2. The method according to claim 1, characterized in that The training process of the network model includes: Constructing a diffusion model and an initial network model, determining noise parameters of the diffusion model in a forward process, and generating noisy channel parameters based on the noise parameters; Determining a first loss function based on the noise parameter and the noisy channel parameter; Inputting the noise channel parameters into the initial network model, and training and adjusting the diffusion model based on a minimization result obtained in the process of minimizing the first loss function to obtain a trained initial network model; Performing progressive distillation training on the trained initial network model to obtain the network model.

3. The method according to claim 2, characterized in that The determining of the noise channel parameters of the diffusion model includes: Obtaining initial channel parameters of the diffusion model in the forward process; The initial channel parameters are subjected to noise addition processing according to a Markov process to obtain the noisy channel parameters of the diffusion model.

4. The method according to claim 2, characterized in that The step of performing progressive distillation training on the trained initial network model to obtain the network model includes: Obtain training channel parameters and the trained initial network model The noise parameters outputted by the number of rounds of processing are used, and a student model corresponding to the trained initial network model is constructed; Based on the training channel parameters and the trained initial network model The noise parameter output by the round processing times determines the second loss function; The training channel parameters are input into the student model, and the student model is trained and adjusted based on the minimization result obtained in the process of minimizing the second loss function to obtain a network model.

5. A channel parameter estimation device, characterized in that: The device is provided in a channel parameter estimation system, the system including a user terminal, an intelligent reflecting surface and a base station in communication connection, and the device includes: An acquisition module is configured to, in response to the base station receiving a transmission signal, wherein the transmission signal is a signal formed by the intelligent reflection-oriented transmission based on the initial signal sent by the user terminal, acquire a corresponding Channel parameters used to estimate the target number of rounds of processing; An input module is configured to input the target estimation channel parameters into a pre-trained network model to obtain noise parameters; The sampling inference module is configured to use the transmission signal, the target estimation channel parameters, the noise parameters and the preset target smart reflector phase to process the sampling inference algorithm to obtain the smart reflector phase. Channel parameters of the round processing times; in response to the When the number of rounds of processing reaches the preset processing number threshold, the The channel parameters of the round processing times are used as the target channel parameters of the smart reflective surface; If the number of rounds of processing does not reach the preset processing number threshold, the The channel parameters of the round processing times are used as the target estimation channel parameters, based on the transmission signal, the The channel parameters of the number of rounds of processing and the phase of the target intelligent reflector are subjected to an iterative reasoning process, and the iterative reasoning process is as follows: An iterative reasoning module is configured to The channel parameters of the round processing times and the phase of the target intelligent reflector are processed by the phase update algorithm to obtain The phase of the intelligent reflector with the number of rounds of processing; The phase of the smart reflector with the highest number of rounds of processing is taken as the target phase of the smart reflector. The channel parameters of the number of rounds of processing are used as target estimation channel parameters, and the transmission signal, the The channel parameters of the number of rounds of processing, the noise parameters and the The phase of the smart reflector with the number of rounds of processing is processed by the sampling inference algorithm to obtain the phase of the smart reflector at Channel parameters for the number of rounds of processing; The repetitive execution module is configured to respond to the If the number of rounds of processing does not reach a preset processing number threshold, the iterative reasoning process is repeated until the channel parameters of the latest round of processing are obtained, and the latest round of processing reaches the preset processing number threshold, and the channel parameters of the latest round of processing are used as the target channel parameters of the smart reflecting surface; The transmission signal, the target estimation channel parameter, the noise parameter and the preset target smart reflector phase are processed by a sampling inference algorithm to obtain the smart reflector phase. Channel parameters for round processing times, including: The phase of the smart reflector surface is determined by the following formula according to the transmission signal, the target estimation channel parameter, the noise parameter and the preset target smart reflector surface phase. Channel parameters for round processing times: in, Indicates that the smart reflective surface is Channel parameters for the number of rounds of processing, , , , represents the variance, , represents the channel parameters used for target estimation, represents the noise parameter, represents the preset first empirical parameter, Indicates the phase of the target intelligent reflector surface, Indicates the transmission signal, Obeying the standard normal distribution, represents the conjugate transpose; The transmission signal, the The channel parameters of the round processing times and the phase of the target intelligent reflector are processed by the phase update algorithm to obtain Smart reflector phase with multiple rounds of processing, including: Based on the transmission signal, the The channel parameters of the round processing times and the phase of the target intelligent reflector are determined by the following formula Smart reflector phase with round processing times: in, express Smart reflector phase with round processing times, Represents the preset second empirical parameter, represents the channel parameters used for target estimation, represents the conjugate transpose, Indicates the transmission signal, Indicates the phase of the target intelligent reflector surface; said utilizing said transmission signal, said The channel parameters of the number of rounds of processing, the noise parameters and the The phase of the smart reflector with the number of rounds of processing is processed by the sampling inference algorithm to obtain the phase of the smart reflector at Channel parameters for round processing times, including: Using the transmission signal, the The channel parameters of the number of rounds of processing, the noise parameters and the The phase of the smart reflector after the number of rounds of processing is determined by the following formula: Channel parameters for round processing times: in, Indicates that the smart reflective surface is Channel parameters for the number of rounds of processing, , , , represents the variance, , Indicates that the smart reflective surface is Channel parameters for the number of rounds of processing, represents the noise parameter, represents the preset first empirical parameter, express Smart reflector phase with round processing times, Indicates the transmission signal, Obeying the standard normal distribution, represents the conjugate transpose.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 4 is implemented.

7. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 4.

Images