A channel estimation method for an OTFS modulation communication system

Abstract

The application discloses a channel estimation method of an OTFS modulation communication system and belongs to the technical field of wireless communication. In view of the problems that there are many parameters to be estimated and high cost in the prior art when channel support set acquisition and channel estimation are carried out, the method firstly starts from three-dimensional joint sparsity of a time-delay-doppler-angle domain, converts the channel estimation problem into an optimization problem of conditional expectation maximization through fitting of a statistical channel model, iteratively optimizes hyperparameters to obtain a channel support set of the time-delay-doppler-angle domain, then based on this, uses an orthogonal search traversal method to screen a channel support set based on a time-delay domain, determines the number of propagation paths between a user and an AP antenna, obtains a doppler-angle domain channel support set of each propagation path, and estimates a channel on the composed three-dimensional channel support set. The method reduces pilot cost through pre-learning, improves the accuracy of OTFS channel estimation in a high-speed mobile scene and the performance of a communication system.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, specifically referring to a channel estimation method for an OTFS (Orthogonal Time Frequency Space) modulation communication system. Background Technology

[0002] OTFS modulation, a two-dimensional modulation method, uses a novel carrier waveform in the delay-Doppler domain to achieve multiplexing of information symbols, finding common ground between traditional time-domain modulation (TDM) and frequency-domain modulation (FDM). More broadly, OTFS combines technologies from the radar and communications fields. Changes in the OTFS waveform due to the influence of the wireless channel directly reflect the physical environment, allowing the derivation of high-resolution radar images of reflectors in the transmission environment that reflect wireless signals. In the future, B5G / 6G will endow next-generation wireless communication networks with more advanced communication and sensing performance, effectively supporting innovative applications such as intelligent driving and intelligent transportation vehicle-to-everything (V2X) networks. However, the high Doppler effect caused by high-speed vehicle movement significantly increases inter-carrier interference and pilot overhead in existing Orthogonal Frequency Division Multiplexing (OFDM) systems, especially exacerbated by the widespread application of millimeter waves and terahertz frequencies in the B5G / 6G era. OTFS technology has attracted widespread attention in the industry due to its significant advantages in resisting time-frequency dual-domain selective fading. Channel estimation techniques based on OTFS have become a research hotspot in recent years.

[0003] Channel estimation is a core technology in communication systems because the wireless channel has a significant impact on system performance. Unlike the fixed nature of wired channels, wireless channels are highly uncertain. They are also greatly affected by environmental factors, such as large-scale fading due to location and distance, time-varying channel characteristics, and frequency-selective fading caused by multipath effects. Furthermore, the increasing maturity of vehicular and aerial communication technologies necessitates consideration of high-speed user mobility in wireless communication systems, which exacerbates the time-varying characteristics and Doppler shift of the wireless channel, leading to additional frequency or phase shifts in the transmitted signal. Considering all these factors, the channel path between the receiver and transmitter becomes highly complex, and the wireless communication channel between them often exhibits significant randomness, posing a considerable challenge to receiver design. To help the receiver better detect and receive signals, wireless channel estimation is necessary. The accuracy of the channel estimation directly affects the performance of the communication system; therefore, channel estimation is a crucial technology in wireless communication systems.

[0004] Chinese patent application CN113852579A, published on December 28, 2021, discloses a low-dimensional subspace OTFS channel estimation method. It proposes a low-dimensional subspace modeling method to model the channel of an OTFS system. This modeling mechanism significantly reduces the number of unknown responses in the Equivalent Channel Response (ECR), thus transforming channel estimation into ECR projection coefficient estimation within the subspace. Furthermore, due to the small number of projection coefficients in the subspace, the originally underdetermined channel estimation problem can be transformed into an overdetermined least-squares coefficient solution. Specifically, a time-domain orthogonal space is constructed using a set of basis functions, and the channel impulse response (CIR) is fitted using these basis functions. By transforming the time-domain basis functions into time-delay-Doppler domain basis functions, the conversion from CIR to ECR is achieved. The projection coefficients of the ECR in the low-dimensional subspace are then obtained by solving for the pilot signals, allowing the ECR to be reconstructed using the basis functions and projection coefficients of the DD domain subspace. However, this technical solution has the following problems:

[0005] (1) Existing technologies use basis functions to estimate subspace equivalent channel parameters. Due to the large number of basis function parameters that need to be estimated for high mobility, this technology requires high pilot overhead. Especially for massive MIMO systems, pilot overhead is particularly important. Therefore, this approach is not suitable for OTFS modulation channel estimation in massive MIMO systems.

[0006] (2) Existing technical solutions rely on different time-domain subspace basis functions to distinguish channel state information and estimate channel coefficients by estimating the parameters of the basis functions, thereby reconstructing the channel. However, when there are many antennas or users in the system, time-domain subspace overlap may occur, meaning that the time-domain basis functions may become non-orthogonal. It is difficult to distinguish antenna ports in more complex systems using only time-domain basis functions, and accurate channel estimation cannot be achieved. Summary of the Invention

[0007] To address the shortcomings of existing technologies that use orthogonal basis functions in the time-delay domain subspace for channel support set acquisition and channel estimation, which involve a large number of parameters to be estimated and high overhead, this invention provides a channel estimation method for OTFS modulated communication systems. By pre-learning, the method reduces pilot overhead, improves the accuracy of OTFS channel estimation in high-speed mobile scenarios, and enhances the performance of the communication system.

[0008] The present invention provides a channel estimation method for an OTFS modulated communication system, comprising the following steps:

[0009] Step 1: Starting from the three-dimensional joint sparsity of the time-delay-Doppler-angle domain, the channel estimation problem is transformed into an optimization problem of conditional expectation maximization by fitting a statistical channel model. The optimization problem is solved, and the hyperparameters are iteratively optimized to obtain a closed-form solution of the hyperparameter update expression. This is to better capture the three-dimensional sparsity characteristics of the time-delay-Doppler-angle domain channel and obtain the channel support set of the time-delay-Doppler-angle domain.

[0010] The optimization problem is described as follows:

[0011] Where, ρ (t+1) Let λ, μ, η represent the hyperparameters of the (t+1)th iteration; λ = {λ n}, λ n Sparsity represents the channel vector element h. n The non-zero probability, h n Let represent the nth term of the channel vector h; μ represents the variance of the complex Gaussian distribution followed by the channel vector h; η represents the noise variance; E{lnp(h,y)|y; ρ (t) Let} denote the conditional expectation of the joint probability density function p(h,y) of the channel vector h and the received signal y.

[0012] Pre-set threshold ∈ 1 and maximum number of iterations T MAX To obtain satisfaction Or it reaches the maximum number of iterations T. MAX A set of hyperparameters for time; N represents the dimension of the channel vector h. The obtained hyperparameters λ are rearranged into a three-dimensional tensor, thus obtaining the channel support set Λ in the time-delay-Doppler-angular domain.

[0013] Step 2: Based on the obtained channel support set Λ, estimate the channel state using an orthogonal search traversal method.

[0014] First, a channel support set Ω is searched in the channel delay domain based on the channel support set Λ, and then filtered based on the delay domain. d The number N of propagation paths between the user and the access point (AP) antenna. P And determine the corresponding Doppler-angle domain channel support set Ω DA The index is used to construct a three-dimensional channel support set Ω based on the currently selected time-delay domain channel support set and the corresponding Doppler-angle domain channel support set. The channel estimation result is calculated based on this three-dimensional channel support set. Finally, all time-delay domain indices are traversed to obtain complete channel state information.

[0015] The advantages and positive effects of this invention are as follows:

[0016] (1) In the prior art, only the orthogonal basis functions of the time domain subspace are used as the channel support set, which requires large-scale parameter estimation; while the method of the present invention uses the sparse information of the three-dimensional channel and utilizes the sparse characteristics of the time delay-Doppler-angle domain to obtain the channel support set with higher accuracy and lower pilot overhead through the fitting of the statistical channel model and the iterative optimization of the hyperparameters.

[0017] (2) Existing technologies for sparse signal recovery based on channel support sets determined in the time delay domain cannot achieve accurate channel estimation. The sparse signal recovery channel estimation scheme based on the channel support set acquisition method proposed in this invention performs orthogonal search traversal in three dimensions: time delay domain, Doppler domain, and angular domain. First, it filters the channel support set index based on the time delay domain to determine the corresponding index of the Doppler-angular domain channel support set, and then recovers the channel state information through orthogonal matching pursuit. This invention performs channel estimation based on the sparse signal recovery problem using the channel support set, and with the assistance of the orthogonal matching pursuit algorithm, it can effectively estimate and recover channel parameters.

[0018] (3) Compared with channel estimation schemes that rely solely on orthogonal basis functions in the time-delay domain subspace, the method of this invention improves channel estimation performance and reduces pilot overhead required for channel estimation. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a distributed massive MIMO system in which the present invention is applied;

[0020] Figure 2 This is a flowchart of the method for obtaining the channel support set in this invention;

[0021] Figure 3 This is a flowchart of the method for estimating the channel based on the channel support set in this invention;

[0022] Figure 4 This is a comparison chart of the performance of the method of the present invention and the existing method in two scenarios with the normalized mean square error (NMSE) varying with the signal-to-noise ratio.

[0023] Figure 5 This is a comparison chart of the NMSE performance of the method of the present invention and existing methods under different pilot overheads;

[0024] Figure 6 This is a comparison chart of the bit error rate (BER) performance of the method of this invention and existing methods. Detailed Implementation

[0025] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0026] This invention considers the location information of distributed wireless access points (APs) within a cell system, including the sparsity of distance (corresponding to the delay domain) and angle (corresponding to the Doppler and angular domains). It uses a statistical model within a Bayesian framework to fit the data, transforming the channel estimation problem into a conditional expectation-maximization (EM) optimization problem. The hyperparameters of this optimization problem are then solved to determine the channel support set. Finally, the channel parameters are solved using the sparse signal recovery orthogonal matching pursuit algorithm to obtain complete channel state information. Compared to existing solutions, this invention improves the accuracy of OTFS channel estimation in high-speed mobile scenarios, reduces pilot overhead, and enhances the performance of the communication system.

[0027] The overall architecture of a distributed massive MIMO system used in this invention is as follows: Figure 1 As shown, the system includes a Central Processing Unit (CPU), an Access Point (AP), and User Equipment (UE). The AP can be a base station or a specially placed dedicated relay device, and it deploys a linear antenna array. The UE can be any access device, including but not limited to mobile phones, computers, and smart tablets, and can also be deployed on highly mobile vehicles such as high-speed trains, unmanned vehicles, and drones. The CPU is the baseband signal processing device in the system, responsible for calculating the signals received by the AP and performing channel estimation and reception detection. Assume that N antennas are deployed on the AP. T There are N transmitting antennas, and the time-varying channel between the user and the AP antenna contains N. P The dominant transmission path.

[0028] This invention realizes a channel estimation method for an OTFS modulation communication system with low pilot overhead, which mainly includes: (1) obtaining the channel support set by optimizing hyperparameters based on the expectation-maximization algorithm; (2) estimating and recovering the channel parameters with the assistance of the orthogonal matching pursuit algorithm.

[0029] like Figure 2 As shown, this invention starts from the three-dimensional joint sparsity of the time-delay-Doppler-angle domain. By fitting a statistical channel model, the channel estimation problem is transformed into a conditional expectation-maximization optimization problem. The expectation-maximization algorithm can clarify the optimization problem and hyperparameter update rules. With the support of the nearest neighbor algorithm, the closed-form solution of the hyperparameter update expression is obtained through iterative updating of hyperparameters, which better captures the three-dimensional sparsity characteristics of the time-delay-Doppler-angle domain channel, thereby realizing a channel estimation scheme that reduces pilot overhead through pre-learning.

[0030] First, the channel estimation problem is transformed into an optimization problem of maximizing conditional expectation, as described below:

[0031]

[0032] Where ρ represents the hyperparameters to be updated, including λ, μ, η, and λ = {λ n}, λ n Sparsity, i.e., the channel vector element h n The non-zero probability, h n Let represent the nth term of the channel vector h, μ represent the variance of the complex Gaussian distribution followed by the channel vector, and η represent the noise variance; t represents the t-th iteration; E{lnp(h,y)|y; ρ (t) Let} represent the conditional expectation of the joint probability density function p(h,y) of the channel vector h and the received signal y. The dimension of the channel h is N = N0. l N k N T =MN T N l and N k These represent the number of resource units along the time-delay domain and along the Doppler domain, respectively.

[0033] Then, the received signal y and pilot matrix Φ are obtained, and the hyperparameters are optimized using the expectation-maximization algorithm to obtain the channel support set, including the following steps 11 to 14.

[0034] Step 11: Parameter initialization, including: setting the initial iteration count t = 1, and setting intermediate parameters. Simultaneously set the maximum number of iterations T. MAX Initial hyperparameter ρ (0) Set the threshold to 1, and set the mean of the initial channel vector. and the variance of the channel vector ρ (0) Include μ (0) and η (0) .

[0035] Step 12: Update the parameters according to the expectation-maximization algorithm, including:

[0036] (a) Update the intermediate parameters of the t-th iteration as follows:

[0037]

[0038]

[0039] in, These are all intermediate process parameters from the t-th iteration and have no practical significance; Φ mn This represents the element in the m-th row and n-th column of the pilot matrix Φ; ym η represents the m-th element of the received signal vector y. (t-1) This represents the noise variance in the (t-1)th iteration. Let represent the mean of the channel vector in the (t-1)th iteration. Let represent the variance of the channel vector in the (t-1)th iteration.

[0040] (b) Update the intermediate variable parameters for the t-th iteration as follows:

[0041]

[0042]

[0043] in, All of these are intermediate variable parameters from the t-th iteration.

[0044] (c) Update the statistical parameters as follows:

[0045]

[0046]

[0047]

[0048] in, Let represent the sparsity of the update in the t-th iteration. ζ n This represents the normalization constant; here, β... n γ n The latest intermediate variable parameter in the current iteration, i.e., the one updated above. Represent a complex Gaussian distribution; Let represent the mean and variance of the channel vector in the current t-th iteration, respectively. In this embodiment of the invention, the normalization constant ζ n Related to the current intermediate parameters,

[0049] (d) Update hyperparameter ρ (t) ,as follows:

[0050]

[0051]

[0052]

[0053] Where, ρ (t) Represents the hyperparameters of the t-th iteration, including μ (t) η (t)Γ(n,a) is defined as h in the channel vector h n ξ is the set of indices of all near-nearest neighbor elements of the a-th quasi-neighbor; n,a Let |Γ(n,a)| represent the weights of Γ(n,a); |Γ(n,a)| represents the number of elements in Γ(n,a). This represents the element corresponding to the b-th term in Γ(n,a) during the t-th iteration.

[0054] When obtaining the set Γ(n,a), a quasi-neighbor is defined as the union of ordinary neighbor nodes and cyclic modulo neighbor nodes. Due to the burst sparsity of angular domains, there are no boundary nodes in angular domains, and the indices of neighbor nodes are determined by cyclic modulo. However, Doppler domains have boundary nodes, so neighbor nodes are directly determined by ordinary neighbors. This invention proposes λ... n The update rule incorporates the sparse characteristics of the channel, and through multiple iterations, it can better capture the channel support set of the sparse channel.

[0055] Step 13: By setting an appropriate threshold ∈1 for λ, the sparse characteristics of the time-delay-Doppler-angle domain channel can be captured through hyperparameters.

[0056] Through multiple iterations of the algorithm, the condition that best satisfies the preset threshold can be found.

[0057] Or t = T MAX A set of hyperparameters, where ||.||2 is the L2 norm.

[0058] Step 14: Rearrange the thresholded λ into a three-dimensional tensor Λ, which is the obtained channel support set in the time-delay-Doppler-angle domain. The following will combine... Figure 3 This demonstrates how the channel state is estimated based on the obtained channel support set.

[0059] This invention transforms the channel estimation problem into a sparse signal recovery problem, such as... Figure 3 As shown, based on the channel support set Λ obtained above in the time-delay-Doppler-angle domain, the channel state is estimated using an orthogonal search traversal method. First, the channel support set based on the time-delay domain is selected, and the index of the corresponding Doppler-angle domain channel support set is determined. Then, a three-dimensional channel support set is constructed based on the currently selected time-delay domain channel support set and its corresponding Doppler-angle domain channel support set, and the channel estimation result is calculated based on this. Finally, all time-delay domain indices are traversed to obtain complete channel state information. This embodiment of the invention uses an orthogonal matching pursuit algorithm based on the channel support set to obtain the three-dimensional channel support set and estimate the channel, including the following steps 21-24.

[0060] Step 21: Set initial parameters, including: algorithm counter i = 1, channel support set. Channel support set in the delay domain Channel support set in Doppler-angle domain Set the initial multipath number N. P =0, set the threshold ∈2. First, based on Λ, filter the indices contained in the channel support set in the delay domain to obtain channel multipath information and delay information.

[0061] Search the channel support set Λ in the channel delay domain. Traverse the channel delay domain indices. If the current channel delay domain index l satisfies the condition ||Λ(l,:)||2>∈2, then update the channel support set Ω in the delay domain. d =Ω d ∪l, update N P Increment by 1. Λ(l,:) represents the item in the channel support set Λ corresponding to the delay domain index l.

[0062] Step 22: Based on the different time delay domain index information, traverse the channel support set Ω of the time delay domain obtained in Step 21. d The channel support set in the Doppler-angle domain is determined by the threshold ∈2.

[0063] For the i-th propagation path, if the following conditions are met... Then update the channel support set of the Doppler-angle domain for the i-th path. Where k and r represent the indices in the channel Doppler domain and the channel angular domain, respectively. Represents the set Ω d The i-th channel delay domain index; i = 1, 2, ..., N P . The corresponding channel support set Λ Element.

[0064] Step 23: Take the Ω obtained in step 21 d The index and the channel support set of each propagation path in the Doppler-angle domain obtained in step 22 The index is incorporated into the set of items to be processed, and the channel support set Ω is updated using the set of items to be processed, as follows:

[0065]

[0066] Channel estimation based on set Ω yields the following channel estimation results. Where Φ represents the pilot matrix, (.)| Ω The position of the index is determined by Ω, superscript. h represents the pseudo-inverse of a matrix. (i) This represents the channel vector of the i-th path.

[0067] Step 24: Determine whether to traverse the entire time-delay domain channel support set, i.e., determine whether i is greater than N. PIf so, it means that the entire time-delay domain channel support set has been traversed, and the final channel state information h = {h (i) i = 1, 2, ..., N P Otherwise, update the value of i, increment i by 1, and repeat steps 22-23.

[0068] The computational complexity of the method in this invention mainly focuses on two parts: one is the hyperparameter learning part, where the computational complexity is measured by the scale of parameter updates and the number of iterations since each parameter is updated in scalar form; the other part is the pseudo-inverse calculation of the pilot matrix in sparse signal recovery. The total complexity can be expressed as follows: Where D MAX T is the index corresponding to the maximum Doppler shift. MAX N P They are usually much smaller than other parameters.

[0069] Two existing methods are selected for comparison. Scheme 1 is a channel estimation scheme based on embedded pilots, and Scheme 2 is a recovery scheme based on sparse signals. The pilot overhead of Scheme 1 is compared with N. l N k N T The guard interval is proportional to the number of pilots, and it also occupies a considerable portion of the spectrum resources. Furthermore, as the number of pilots increases, different pilots will interfere with each other due to insufficient guard intervals, leading to a decrease in channel estimation performance. Therefore, it is not suitable for the application scenario of this invention. The pilot overhead of Scheme 2 is typically proportional to Qlog(N). l N k N T This is the theoretical lower bound for pilot overhead. Although it reduces the pilot overhead from linear to logarithmic, the sparsity of the channel is often a priori unknown in actual communication, such as multipath number and multi-block sparse patterns in the Doppler-angle domain. Therefore, such methods usually need to determine the channel support set through correlation operations, thus requiring a higher pilot overhead than the theoretical lower bound to ensure the accuracy of the estimation. In this invention, the iterative part does not bring a huge increase in complexity. Even though the surge in the number of antennas will bring a linear increase in complexity, by setting an appropriate number of OTFS resource blocks (N)... l N k This allows the complexity to be kept within an acceptable range. Therefore, it can be seen that the method of this invention first accurately estimates the location of the channel support set through parameter learning, and the learning process can be performed offline, which can greatly reduce pilot overhead.

[0070] The following experiments compare the technical effects of the method of this invention with two existing solutions. Two user speeds are set to approximate different scenarios: 120 km / h represents a high-speed driverless car scenario, and 360 km / h represents a high-speed rail scenario. The method of this invention is abbreviated as JSPL, the first solution is abbreviated as IB, and the second solution corresponds to OMP and 3D-SOMP respectively.

[0071] like Figure 4 As shown, the NMSE performance of all schemes is compared under varying signal-to-noise ratio (SNR). All schemes perform better at lower speeds (120 km / h) than at higher speeds (360 km / h). This is because the Doppler shift caused by high speeds is more severe, thus having a greater impact on channel estimation performance. Furthermore, the larger the Doppler shift, the lower the resolution of a single grid point with a fixed number of OTFS resource blocks, making it more difficult to recover the channel support set, thereby affecting the accuracy of channel estimation. Scheme 1 suffers a significant performance drop due to the lack of sufficient guard intervals in massive MIMO systems; while Scheme 2, lacking prior knowledge of channel sparsity, such as multipath number and channel support set, also performs worse than the method of this invention. Moreover, it can be seen that the performance improvement of the method of this invention is greater at lower speeds than at higher speeds, also because the grid resolution decreases with larger Doppler shifts, affecting the accuracy of channel recovery.

[0072] like Figure 5 As shown, with the user's speed set to 360 km / h, the NMSE performance of all schemes was compared under different pilot overheads σ. It can be seen that the method of this invention can save 90% of the pilot overhead compared to other methods, while also having superior estimation accuracy. This is because the method of this invention learns an efficient channel support set recovery, saving the additional pilot overhead of sparse signal recovery algorithms in determining the channel support set through correlation operations, thus reaching the theoretical lower bound of pilot overhead.

[0073] like Figure 6 As shown, the BER performance of the method of this invention is compared with that of existing methods in 4QAM modulation. The BER is also compared when the channel state information (Perfect channel) is fully known. The user speed is set to 360 km / h. As can be seen from the figure, the pilot overhead of the method of this invention is 5%, and the Perfect channel overhead is 0%, while the pilot overhead of the existing scheme is 50%. The method of this invention outperforms the other schemes and has a smaller performance lower bound compared to the Perfect channel.

[0074] Except for the technical features described in the specification, all other technologies are known to those skilled in the art. Descriptions of well-known components and technologies are omitted in this invention to avoid redundancy and unnecessary limitation. The embodiments described above do not represent all embodiments consistent with this application. Various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this invention are still within the protection scope of this invention.

Claims

1. A channel estimation method for an OTFS modulated communication system, characterized in that, Includes the following steps: Step 1: Starting from the three-dimensional joint sparsity of time delay-Doppler-angle domain, the channel estimation problem is transformed into an optimization problem of conditional expectation maximization. Then, the channel support set is obtained by optimizing the hyperparameters based on the expectation maximization algorithm. The optimization problem is described as follows: Where, ρ (t+1) Let λ, μ, η represent the hyperparameters of the (t+1)th iteration; λ = {λ n }, λ n Sparsity represents the channel vector element h. n The non-zero probability, h n Let represent the nth term of the channel vector h; μ represents the variance of the complex Gaussian distribution followed by the channel vector h; η represents the noise variance; E{ln p(h,y)|y; ρ (t) } represents the conditional expectation of the joint probability density function p(h,y) of the channel vector h and the received signal y; Methods for hyperparameter optimization and obtaining channel support sets include: Step 11: Input the received signal y and the pilot matrix Φ; set the initial parameters, including: the initial value of the iteration number t is 1, and the intermediate process parameters. The initial hyperparameter ρ is 0. (0) , ρ (0) Include μ (0) and η (0) Threshold ∈ 1, maximum number of iterations T MAX The mean of the initial channel vector and variance Step 12: Optimize hyperparameters based on the expectation-maximization algorithm, including: (a) Update the intermediate parameters of the t-th iteration as follows: Where, Φ mn It is the element in the m-th row and n-th column of the pilot matrix Φ. and These are the mean and variance of the channel vector in the (t-1)th iteration, respectively. m It is the m-th element of the received signal y, η (t-1) Let represent the noise variance in the (t-1)th iteration; (b) Update the intermediate variable parameters in the t-th iteration. as follows: (c) Update the statistical parameters as follows: in, It is the sparsity of the t-th iteration update. Describes a complex Gaussian distribution; β n γ n Substitute the updated parameters from (b) into each parameter; ζ n Represents the normalization constant; and These are the mean and variance of the channel vector in the t-th iteration, respectively; (d) Update hyperparameter ρ (t) ,Include μ (t) and η (t) ,as follows: Where Γ(n,a) is defined as the set of indices of all quasi-neighbor elements of the nth term in the channel vector h, and ξ n,a Let |Γ(n,a)| represent the weights of Γ(n,a), and |Γ(n,a)| represent the number of elements in Γ(n,a). This represents the element corresponding to the b-th term in Γ(n,a) during the t-th iteration. M = N l N k N l and N k These represent the number of resource units along the time-delay domain and along the Doppler domain, respectively. Step 13: Pre-set the threshold ∈1 and the maximum number of iterations, iteratively execute step 12, and obtain the results that satisfy the condition. Alternatively, a set of hyperparameters that reaches the maximum number of iterations; N represents the dimension of the channel vector h; rearrange λ in the obtained hyperparameters into a three-dimensional tensor to obtain the channel support set Λ; Step 2: Estimate the channel state using orthogonal search traversal based on the obtained channel support set Λ; First, based on the channel support set Λ, a search is performed in the channel delay domain to obtain the channel support set Ω based on the delay domain. d The number N of propagation paths between the user and the AP antenna P And determine the corresponding Doppler-angle domain channel support set Ω DA Then, based on the currently selected time-delay domain channel support set and the determined Doppler-angle domain channel support set, a three-dimensional channel support set Ω is constructed. The channel is estimated based on the three-dimensional channel support set Ω, and finally, the complete channel state is obtained by traversing all the time-delay domain indices. The filtering process yields the channel support set Ω based on the time delay domain. d This includes: setting a threshold ∈ 2, searching the channel support set Λ in the channel delay domain, and if the current channel delay domain index... Meet the conditions Then the index Merge into Ω d In, and update N P Increment by 1; for the i-th propagation path, if... Then (k,r) will be merged into the channel support set of the Doppler-angle domain of the i-th path. In the diagram, k and r represent the indices of the channel Doppler domain and the channel angular domain, respectively. It is a set Ω d The channel delay domain index.

2. The method according to claim 1, characterized in that, In step 2, the channel is estimated based on the three-dimensional channel support set Ω, and the channel vector of the i-th propagation path is estimated. Where Φ represents the pilot matrix, (.)| Ω The position of the index is determined by Ω, superscript. Represent the pseudo-inverse of the matrix; traverse all time-delay domain indices to obtain the final channel state h = {h (i) i = 1, 2, ..., N P }

Images