Spatial non-stationary large-scale MIMO channel estimation method based on pre-screening and multistage dynamic threshold strategy

Abstract

The invention discloses a spatial non-stationary large-scale MIMO channel estimation method based on pre-screening and a multi-stage dynamic threshold strategy, belongs to the technical field of wireless communication, and solves the technical problems of insufficient channel estimation precision and relatively high calculation complexity of a spatial non-stationary large-scale MIMO system in the prior art. The channel estimation method is divided into two stages: pre-screening strategy stage estimation: carrying out preliminary screening on all possible candidate atoms so as to reduce the calculation amount; by calculating the correlation between the candidate atoms and the residual error, selecting the candidate atoms which are most likely to be non-zero to form a primary candidate set; estimating in a multi-stage dynamic threshold adjustment stage, further screening candidate atom sets selected according to a pre-selection strategy, and improving the accuracy of non-zero element estimation through a dynamic threshold mechanism; the dynamic threshold is adjusted according to the current signal-to-noise ratio (SNR) so as to adaptively process the estimation precision under different channel conditions and ensure the robustness of channel estimation under different signal-to-noise ratio conditions.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, in particular to a spatial non-stationary large-scale MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy. Background Art

[0002] With the development of 5G and future 6G communication technologies, Massive Multiple Input Multiple Output (Massive MIMO) technology is increasingly being used in wireless communication systems. Massive MIMO systems achieve spatial multiplexing through large-scale antenna arrays, thereby greatly improving the system's spectrum efficiency, communication capacity, and transmission reliability. Due to these advantages, massive MIMO is considered one of the key technologies for future communication systems.

[0003] However, as the size of base station antenna arrays increases, channel models and channel estimation methods face new challenges, especially in spatially non-stationary environments. Spatially non-stationary channels arise due to the large physical size of the antenna array and the different channel environments experienced by different parts of the array, which causes the signals received by different antennas to have different statistical properties. Specifically, different parts of the antenna array may be exposed to different scattering clusters, which causes the channel characteristics between antennas to exhibit spatial non-stationarity. Therefore, traditional channel estimation methods, such as least squares (LS) and minimum mean square error (MMSE), cannot accurately estimate the channel state when faced with such complex non-stationary channels, resulting in performance degradation.

[0004] Most existing channel estimation methods assume that the channel has a uniform sparse structure. However, in spatially non-stationary channels, the sparsity of the channel usually manifests itself as a non-uniform distribution, resulting in poor performance of traditional estimation algorithms in this scenario. Commonly used compressed sensing methods, such as orthogonal matching pursuit (OMP) and sparse adaptive matching pursuit (SAMP), perform poorly when dealing with non-uniform sparse channels and it is difficult to obtain high-precision channel estimates. In addition, although some methods such as sparse Bayesian learning (SBL) improve the estimation accuracy, their high computational complexity limits their application in actual large-scale systems.

[0005] Faced with the problem of spatial non-stationary channels in large-scale MIMO systems, there is an urgent need for an efficient and robust channel estimation method that can accurately estimate the non-uniform sparse structure in spatial non-stationary channels while keeping the computational complexity low. Summary of the invention

[0006] In view of the shortcomings and deficiencies in the prior art, the present invention provides a hierarchical channel estimation method based on pre-selection and multi-level dynamic threshold strategies. By dividing the channel estimation process into two levels and combining the pre-selection strategy and dynamic threshold adjustment, the channel sparse structure in a spatial non-stationary environment can be adaptively estimated, and adaptive processing of channel sparsity and non-uniformity is achieved, thereby significantly improving the accuracy and efficiency of channel estimation. The spatial non-stationary large-scale MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategies.

[0007] This method processes the channel estimation process in layers.

[0008] To achieve the above objectives, the present invention is implemented by the following technical scheme: The spatial non-stationary large-scale MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy provided by the present invention comprises the following steps:

[0009] S1 Establishment of system model;

[0010] The S2 hierarchical channel estimation method divides the channel estimation process into two stages: pre-screening strategy stage estimation and multi-level dynamic threshold adjustment stage estimation;

[0011] S2.1 Pre-screening strategy stage estimation, firstly, preliminarily screen all possible candidate atoms to reduce the amount of calculation; the main purpose of this estimation stage is to select the candidate atoms that are most likely to be non-zero by calculating the correlation between the candidate atoms and the residuals to form the preliminary candidate set;

[0012] S2.2 Multi-level dynamic threshold adjustment stage estimation, further screening the candidate atom set selected by the pre-selection strategy, and improving the accuracy of non-zero element estimation through the dynamic threshold mechanism; the dynamic threshold is adjusted according to the current signal-to-noise ratio (SNR) to adaptively handle the estimation accuracy under different channel conditions, ensuring the robustness of channel estimation under different signal-to-noise ratio conditions;

[0013] S3 obtains and outputs a final channel estimation result through a two-stage channel estimation process of the pre-screening strategy stage estimation and the multi-stage dynamic threshold adjustment stage estimation in step S2.

[0014] Preferably, step S1 of establishing the system model specifically includes the following:

[0015] This method is applicable to large-scale MIMO systems, assuming that the system uses orthogonal frequency division multiplexing (OFDM) technology for signal transmission; the base station has a large-scale antenna array and receives signals from multiple users. The channel sparsity is reflected in the fact that some antennas receive valid signals;

[0016] The received signal is represented as:

[0017] Y=SX+W (1)

[0018] Among them, Y is the received signal matrix, S is the sensing matrix, X is the channel matrix, and W is the noise matrix.

[0019] Preferably, in the hierarchical channel estimation method in step 2, the first stage estimation based on the pre-screening strategy, 2.1 pre-screening strategy stage estimation, the goal of this process is to reduce the computational complexity in channel estimation, and to achieve more accurate and efficient channel estimation by screening potential non-zero atom candidate sets; the specific steps include the following:

[0020] Initialization: Initialize the channel estimation matrix is a zero matrix, the residual matrix R(0) is the received signal matrix Y, and the initial support set Ω(0) = φ;

[0021] (1) Calculate the correlation: Calculate the correlation between each candidate atom and the current residual. The residual matrix R is expressed as:

[0022]

[0023] in, is the channel estimation value at the tth iteration;

[0024] (2) Selecting the preliminary candidate set: Pre-select the atoms with the highest correlation according to the correlation ranking to form the preliminary candidate set. The correlation is calculated by the following formula:

[0025]

[0026] Among them, S r is the rth column of the perception matrix, and R is the current residual matrix;

[0027] (3) Update the candidate support set: The selected preliminary candidate set is used to update the current support set. The updated support set is:

[0028] Ψ pre =Ω old ∪Top_Atoms (4)

[0029] Among them, Ω old is the support set in the previous iteration, and Top_Atoms is the candidate atom set obtained through pre-selection.

[0030] Preferably, it also includes:

[0031] (4) Channel estimation update: Use the updated support set to re-estimate the channel matrix And update the residual:

[0032]

[0033] in is the pseudo-inverse of the sensing matrix corresponding to the support set, and Y is the received signal matrix;

[0034] (5) Stop condition: If the preset number of iterations is reached or the residual is less than the set threshold, the iteration ends and the estimated channel matrix is ​​output

[0035] Preferably, the pre-selection strategy in step (2) is: according to the atomic correlation ranking, the first num_pre atoms are screened out to enter the preliminary candidate set; the pre-selection strategy determines the number of screened atoms by a preset ratio, which is set to 10% of the total number of candidate atoms, that is:

[0036] num_pre=max(1,(ratio×n)) (6)

[0037] Where n is the total number of atoms.

[0038] Preferably, in the second stage estimation based on multi-level dynamic threshold modulation in the hierarchical channel estimation method in step 2, 2.2 multi-dynamic threshold adjustment stage estimation, the method accurately estimates the non-zero elements by dynamically adjusting the threshold to ensure the robustness of the channel estimation under different signal-to-noise ratio conditions, and the specific steps include the following:

[0039] Initialization: Use the channel matrix estimated by the first stage As input, initialize the multi-level dynamic threshold parameter λ;

[0040] (1) Calculate the signal-to-noise ratio: First, calculate the signal-to-noise ratio based on the current received signal Y and the noise power σ 2 Calculate the current signal-to-noise ratio SNR current :

[0041]

[0042] Where W is the noise matrix;

[0043] (2) Dynamic threshold adjustment: The dynamic threshold adjustment factor κ is calculated based on the current signal-to-noise ratio and the target signal-to-noise ratio. The adjustment factor κ is calculated using the following formula:

[0044]

[0045] The subscript current in the formula represents the current signal-to-noise ratio, and target represents the given target signal-to-noise ratio. This factor is used to adjust the dynamic threshold under different SNR conditions to ensure better noise suppression under low SNR conditions and improve channel estimation accuracy under high SNR conditions.

[0046] And dynamically modulate the threshold λ to adapt to the current channel conditions:

[0047]

[0048] Where N p is the number of non-zero elements, and E is the support set.

[0049] Preferably, it also includes:

[0050] (3) Accurately screen non-zero elements: Use the adjusted threshold λ to further screen the non-zero elements estimated in the first stage, remove the pseudo non-zero elements introduced by noise interference, and update the support set and estimated value;

[0051] (4) Result output: After all iterations are completed, the final channel estimation matrix is ​​output

[0052] Preferably, the method further comprises, in step S3, the specific steps comprising:

[0053] Through the two-stage channel estimation process, the final channel estimation result is obtained, which is expressed as:

[0054]

[0055] Among them, λ is the estimated parameter adjusted according to the dynamic threshold.

[0056] Preferably, the method further includes simulation verification in step S4, in which the performance of the estimation method is verified through simulation analysis, and specific performance indicators include the number of atom selections, mean square error (NMSE) and reconstruction success probability.

[0057] The present invention provides a spatial non-stationary large-scale MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy. It has the following beneficial effects:

[0058] (1) The spatial non-stationary large-scale MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy of the present invention can adaptively estimate the channel sparse structure in a spatial non-stationary environment by hierarchically processing the channel estimation process, combining the pre-selection strategy and dynamic threshold adjustment, significantly improving the efficiency and accuracy of channel estimation, achieving efficient and accurate estimation of the channel, effectively solving the spatial non-stationary channel problem in large-scale MIMO systems, and reducing the performance degradation problem encountered by traditional compressed sensing algorithms when processing non-uniform sparse structures. The method is suitable for complex wireless communication scenarios.

[0059] (2) Compared with the traditional channel estimation algorithm, the present invention has significant advantages in the following aspects: Reduced computational complexity: Through the pre-selection strategy, the number of atoms that need to be processed is reduced in the initial screening of candidate atoms, thereby reducing the computational complexity.

[0060] Enhanced adaptability: The dynamic threshold adjustment mechanism adaptively adjusts the estimation parameters according to the current signal-to-noise ratio, so that the method can maintain high robustness under different channel conditions.

[0061] Improved estimation accuracy: This method can more accurately capture the non-uniform sparse structure in spatial non-stationary channels, significantly improving the accuracy of channel estimation, especially with better noise resistance in low signal-to-noise ratio scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a comparison chart of the advantages of the method of the present invention in terms of the number of atom selections.

[0063] Figure 2 This is a comparison chart of mean square error and signal-to-noise ratio performance between the method of the present invention and other methods.

[0064] Figure 3 This is a comparison chart of the reconstruction success probability of the method of the present invention and other methods. DETAILED DESCRIPTION

[0065] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0066] Example 1

[0067] The spatial non-stationary large-scale MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy provided by the present invention includes two stages:

[0068] Example 1

[0069] The spatial non-stationary large-scale MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy of the present invention is characterized by comprising the following steps:

[0070] S1 System model establishment; This method is applicable to large-scale MIMO systems, assuming that the system uses orthogonal frequency division multiplexing (OFDM) technology for signal transmission; the base station has a large-scale antenna array and receives signals from multiple users. The channel sparsity is reflected in the fact that some antennas receive valid signals;

[0071] The received signal is represented as:

[0072] Y=SX+W (1)

[0073] Among them, Y is the received signal matrix, S is the sensing matrix, X is the channel matrix, and W is the noise matrix.

[0074] The S2 hierarchical channel estimation method divides the channel estimation process into two stages: pre-screening strategy stage estimation and multi-level dynamic threshold adjustment stage estimation; 2.1 Pre-screening strategy stage estimation, in this estimation stage, all possible candidate atoms are first preliminarily screened to reduce the amount of calculation; the main purpose of this estimation stage is to select the candidate atoms that are most likely to be non-zero by calculating the correlation between the candidate atoms and the residuals to form a preliminary candidate set; 2.2 Multi-level dynamic threshold adjustment stage estimation, further screening is performed on the candidate atom set selected by the pre-selection strategy, and the accuracy of non-zero element estimation is improved through the dynamic threshold mechanism; the dynamic threshold is adjusted according to the current signal-to-noise ratio (SNR) to adaptively process the estimation accuracy under different channel conditions, ensuring the robustness of the channel estimation under different signal-to-noise ratio conditions.

[0075] S3 obtains and outputs a final channel estimation result through a two-stage channel estimation process of the pre-screening strategy stage estimation and the multi-stage dynamic threshold adjustment stage estimation in step S2.

[0076] Example 2

[0077] First-stage estimation based on pre-screening strategy

[0078] 2.1 Pre-screening strategy stage estimation, the goal of this process is to reduce the computational complexity in channel estimation and achieve more accurate and efficient channel estimation by screening potential non-zero atom candidate sets; the specific steps include the following:

[0079] (1) Initialization: Initialize the channel estimation matrix is a zero matrix, the residual matrix R(0) is the received signal matrix Y, and the initial support set Ω(0) = φ;

[0080] (2) Calculate the correlation between atoms and residuals: Calculate the correlation between each candidate atom and the current residual. The residual matrix R is expressed as:

[0081]

[0082] in, is the channel estimation value at the tth iteration;

[0083] (3) Selecting the preliminary candidate set: Pre-select the atoms with the highest correlation according to the correlation ranking to form the preliminary candidate set. The correlation is calculated by the following formula:

[0084]

[0085] Among them, S r is the rth column of the perception matrix, and R is the current residual matrix;

[0086] In each iteration, the inner product of each column in the perception matrix S and the residual matrix R(t-1) is calculated to obtain the atomic correlation of the current iteration:

[0087]

[0088] The pre-selection strategy is: according to the atomic correlation ranking, the first num_pre atoms are selected to enter the preliminary candidate set; the pre-selection strategy determines the number of atoms selected by the preset ratio ratio, which is set to 10% of the total number of candidate atoms, that is:

[0089] num_pre=max(1,(ratio×n)) (5)

[0090] Where n is the total number of atoms.

[0091] (4) Update candidate support set: The selected preliminary candidate set is used to update the current support set. The updated support set is:

[0092] Ψ pre =Ω old ∪Top_Atoms (6)

[0093] Among them, Ω old is the support set in the previous iteration, and Top_Atoms is the candidate atom set obtained through pre-selection.

[0094] (5) Channel estimation update: Use the updated support set to re-estimate the channel matrix And update the residual:

[0095]

[0096] in is the pseudo-inverse of the sensing matrix corresponding to the support set, and Y is the received signal matrix;

[0097] (6) Stop condition: If the preset number of iterations is reached or the residual is less than the set threshold, the iteration ends and the estimated channel matrix is ​​output

[0098] Through the pre-selection strategy, the selected candidate atom set reduces the number of unnecessary atoms, thereby reducing the computational complexity. Especially when the signal-to-noise ratio is low, the method exhibits higher accuracy and convergence speed.

[0099] Pre-selection strategy stage: In the initial stage of channel estimation, a pre-selection strategy is used to screen candidate atoms that may contain non-zero elements. By calculating the correlation between the candidate atoms and the current residual, the set of atoms that are most likely to be non-zero elements is selected first, thereby reducing the amount of calculation and speeding up the estimation process.

[0100] Example 3

[0101] Second-stage estimation based on multi-level dynamic threshold modulation

[0102] 2.2 Multiple dynamic threshold adjustment stage estimation, this method accurately estimates the non-zero elements by dynamically adjusting the threshold to ensure the robustness of the channel estimation under different signal-to-noise ratio conditions. The specific steps include the following:

[0103] (1) Initialization: Use the channel matrix estimated by the first stage As input, initialize the multi-level dynamic threshold parameter λ;

[0104] (2) Calculate the signal-to-noise ratio: First, calculate the signal-to-noise ratio based on the current received signal Y and the noise power σ 2 Calculate the current signal-to-noise ratio SNR current :

[0105]

[0106] Where W is the noise matrix;

[0107] (3) Dynamic threshold adjustment: The dynamic threshold adjustment factor κ is calculated based on the current signal-to-noise ratio and the target signal-to-noise ratio. The adjustment factor κ is calculated using the following formula:

[0108]

[0109] The subscript current in the formula represents the current signal-to-noise ratio, and target represents the given target signal-to-noise ratio. This factor is used to adjust the dynamic threshold under different SNR conditions to ensure better noise suppression under low SNR conditions and improve channel estimation accuracy under high SNR conditions.

[0110] And dynamically modulate the threshold λ to adapt to the current channel conditions:

[0111]

[0112] Where N p is the number of non-zero elements, and E is the support set.

[0113] (4) Accurately screen non-zero elements: Use the adjusted threshold λ to further screen the non-zero elements estimated in the first stage, remove the pseudo non-zero elements introduced by noise interference, and update the support set and estimated value;

[0114] (5) Result output: After all iterations are completed, the final channel estimation matrix is ​​output

[0115] By dynamically adjusting the threshold, this method can adaptively adjust the estimation parameters under different signal-to-noise ratio conditions to ensure the accuracy and robustness of channel estimation.

[0116] S3 obtains the final channel estimation result through the two-stage channel estimation process of the pre-screening strategy stage estimation and the multi-level dynamic threshold adjustment stage estimation, which is expressed as:

[0117]

[0118] Among them, λ is the estimated parameter adjusted according to the dynamic threshold.

[0119] Multi-level dynamic threshold adjustment stage: In the second stage, a dynamic threshold adjustment mechanism is used to dynamically adjust the reconstruction noise parameters according to the current signal-to-noise ratio (SNR) to further accurately estimate the non-zero elements selected in the pre-selection stage. This threshold adjustment mechanism enables the algorithm to adapt to different channel conditions and maintain high robustness in both high SNR and low SNR scenarios.

[0120] Example 4

[0121] S4 Simulation Verification: The performance of the estimation method is verified through simulation analysis. The specific performance indicators include the number of selected atoms, mean square error (NMSE) and reconstruction success probability. The results show that the estimation method of the present invention is superior to traditional algorithms such as OMP and SAMP in terms of performance indicators such as the number of selected atoms, mean square error (NMSE) and reconstruction success probability, especially showing higher robustness under low signal-to-noise ratio conditions.

[0122] Depend on Figure 1 The results of the comparison diagram of the advantages of the method of the present invention in the number of atom selections show that the pre-screening strategy of the present invention requires fewer atoms than the atom selection in the classical hierarchical estimation method. The pre-screening strategy in the present invention effectively filters out unnecessary atoms, improves the accuracy of atom selection, and reduces the computational load. Under low signal-to-noise ratio conditions, the pre-selection strategy in the method of the present invention shows significant advantages in accuracy and convergence speed, ensuring the effectiveness of the overall algorithm. In the algorithm proposed by the present invention, although the added calculation steps slightly increase the complexity, it brings better performance and robustness, especially adapting to noisy environments.

[0123] Figure 2 A comparison chart of mean square error and signal-to-noise ratio performance between the method of the present invention and other methods is given. Figure 2 The mean square error performance analysis in clearly shows that the method of the present invention is always superior to the parallel SAMP, threshold enhanced hierarchical estimation method and adaptive threshold enhanced hierarchical estimation method at all SNR values.

[0124] Under a signal-to-noise ratio of 10dB, the algorithm proposed in the present invention achieves an improvement of about 30% in mean square error, which is 30% less than the threshold-enhanced hierarchical estimation method and 20% less than the adaptive threshold-enhanced hierarchical estimation method. This significant improvement is attributed to the pre-screening strategy of the first-stage estimation and the multi-level dynamic threshold adjustment of the second-stage estimation, which comprehensively improves the accuracy of non-zero element estimation and support set identification.

[0125] In the high SNR range of 20-30 dB, the proposed algorithm is close to the mean square error performance of the benchmark least squares method. Overall, the proposed method shows better mean square error performance, especially at low and medium SNR levels, which verifies the effectiveness and reliability of the proposed estimation method for channel estimation in large-scale MIMO systems.

[0126] Figure 3 A comparison chart of the reconstruction success probability of the method of the present invention and other methods is given. Figure 3 The results show that in terms of reconstruction success probability, the method of the present invention is always better than the parallel SAMP, threshold enhanced hierarchical estimation method and adaptive threshold enhanced hierarchical estimation method at all signal-to-noise ratio values.

[0127] When the signal-to-noise ratio is 10 dB, the algorithm proposed in the present invention achieves a success rate of about 80%, while the success rates of the threshold-enhanced hierarchical estimation method and the adaptive threshold-enhanced hierarchical estimation method are 75% and 77%, respectively, which are improved by 5% and 3%, respectively.

[0128] When the signal-to-noise ratio increases to 20 dB, the method of the present invention maintains a success rate of 95%, which is significantly better than the threshold-enhanced hierarchical estimation method and the adaptive threshold-enhanced hierarchical estimation method, which are 93% and 94% respectively.

[0129] When the signal-to-noise ratio is 30 dB, the success rate of the method of the present invention is close to 99%, which is very close to the performance of the benchmark least squares method.

[0130] Overall, the proposed method shows an excellent success rate in sparse reconstruction, especially at low and medium SNR levels, confirming its robustness in noisy environments. The integration of pre-screening strategy and multi-level dynamic threshold adjustment enables the proposed method to effectively identify and retain important signal components, ensuring accurate channel reconstruction under various SNR conditions.

[0131] In summary, the present invention adopts a layered channel estimation method, combines a preselection strategy with a dynamic threshold adjustment mechanism, and can effectively improve the estimation accuracy and efficiency of spatial non-stationary channels in large-scale MIMO systems. It has the advantages of reduced computational complexity, enhanced adaptability, and improved estimation accuracy. This method can more accurately capture the non-uniform sparse structure in spatial non-stationary channels, significantly improve the accuracy of channel estimation, and especially has better noise resistance in low signal-to-noise ratio scenarios.

[0132] On the basis of the above embodiments, the present invention continues to describe in detail the technical features involved therein and the functions and roles played by the technical features in the present invention, so as to help technicians in this field fully understand the technical solution of the present invention and reproduce it.

[0133] Finally, although this specification is described according to implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A spatially non-stationary massive MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy, characterized in that: The steps include: S1 Establishment of system model; The S2 hierarchical channel estimation method divides the channel estimation process into two stages: pre-screening strategy stage estimation and multi-level dynamic threshold adjustment stage estimation; S2.1 Pre-screening strategy stage estimation: In this estimation stage, all possible candidate atoms are first screened to reduce the amount of calculation; by calculating the correlation between the candidate atoms and the residuals, the candidate atoms most likely to be non-zero are selected to form the preliminary candidate set; S2.2 Multi-level dynamic threshold adjustment stage estimation, further screening the candidate atom set selected by the pre-selection strategy, and improving the accuracy of non-zero element estimation through the dynamic threshold mechanism; the dynamic threshold is adjusted according to the current signal-to-noise ratio (SNR) to adaptively handle the estimation accuracy under different channel conditions, ensuring the robustness of channel estimation under different signal-to-noise ratio conditions; S3 obtains and outputs a final channel estimation result through a two-stage channel estimation process of the pre-screening strategy stage estimation and the multi-stage dynamic threshold adjustment stage estimation in step S2.

2. The spatial non-stationary massive MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy according to claim 1 is characterized in that: Step S1: Establishment of the system model, specifically including the following: This method is applicable to large-scale MIMO systems, assuming that the system uses orthogonal frequency division multiplexing (OFDM) technology for signal transmission; the base station has a large-scale antenna array to receive signals from multiple users, and the channel sparsity is reflected in the fact that some antennas receive valid signals; The received signal is represented as: Y=SX+W (1) Among them, Y is the received signal matrix, S is the sensing matrix, X is the channel matrix, and W is the noise matrix.

3. The spatial non-stationary massive MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy according to claim 1, characterized in that: The first stage estimation based on the pre-screening strategy in the hierarchical channel estimation method in step 2, S2.1 pre-screening strategy stage estimation, the goal of this process is to reduce the computational complexity in channel estimation, by screening potential non-zero atom candidate sets to achieve more accurate and efficient channel estimation, the specific steps include the following: Initialization: Initialize the channel estimation matrix is a zero matrix, the residual matrix R(0) is the received signal matrix Y, and the initial support set Ω(0) = φ; (1) Calculate the correlation: Calculate the correlation between each candidate atom and the current residual. The residual matrix R is expressed as: in, is the channel estimation value at the tth iteration; (2) Selecting the preliminary candidate set: Pre-select the atoms with the highest correlation according to the correlation ranking to form the preliminary candidate set. The correlation is calculated by the following formula: Among them, S r is the rth column of the perception matrix, and R is the current residual matrix; (3) Update the candidate support set: The selected preliminary candidate set is used to update the current support set. The updated support set is: P pre =Oh old ∪Top_Atoms (4) Among them, Ω old is the support set in the previous iteration, and Top_Atoms is the candidate atom set obtained through pre-selection.

4. The spatial non-stationary massive MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy according to claim 3, characterized in that: Also includes: (4) Channel estimation update: Use the updated support set to re-estimate the channel matrix And update the residual: in is the pseudo-inverse of the sensing matrix corresponding to the support set, and Y is the received signal matrix; (5) Stop condition: If the preset number of iterations is reached or the residual is less than the set threshold, the iteration ends and the estimated channel matrix is ​​output 5. The spatial non-stationary massive MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy according to claim 3, characterized in that: The pre-selection strategy in step (2) is: according to the atomic correlation ranking, the first num_pre atoms are selected to enter the preliminary candidate set; the pre-selection strategy determines the number of atoms selected by the preset ratio ratio, which is set to 10% of the total number of candidate atoms, that is: num_pre=max(1,(ratio×n)) (6) Where n is the total number of atoms.

6. The spatial non-stationary massive MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy according to claim 2, characterized in that: The second stage estimation based on multi-level dynamic threshold modulation in the hierarchical channel estimation method in step 2, S2.2 multi-dynamic threshold adjustment stage estimation, this method accurately estimates the non-zero elements by dynamically adjusting the threshold to ensure the robustness of the channel estimation under different signal-to-noise ratio conditions, the specific steps include the following: Initialization: Use the channel matrix estimated by the first stage As input, initialize the multi-level dynamic threshold parameter λ; (1) Calculate the signal-to-noise ratio: First, calculate the signal-to-noise ratio based on the current received signal Y and the noise power σ 2 Calculate the current signal-to-noise ratio SNR current : Where W is the noise matrix; (2) Dynamic threshold adjustment: The dynamic threshold adjustment factor κ is calculated based on the current signal-to-noise ratio and the target signal-to-noise ratio. The adjustment factor κ is calculated using the following formula: The subscript current in the formula represents the current signal-to-noise ratio, and target represents the given target signal-to-noise ratio. This factor is used to adjust the dynamic threshold under different SNR conditions to ensure better noise suppression under low SNR conditions and improve channel estimation accuracy under high SNR conditions. And dynamically modulate the threshold λ to adapt to the current channel conditions: Where N p is the number of non-zero elements, and E is the support set.

7. The spatial non-stationary massive MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy according to claim 6, characterized in that: Also includes: (3) Accurately screen non-zero elements: Use the adjusted threshold λ to further screen the non-zero elements estimated in the first stage, remove the pseudo non-zero elements introduced by noise interference, and update the support set and estimated value; (4) Result output: After all iterations are completed, the final channel estimation matrix is ​​output 8. The spatial non-stationary massive MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy according to claim 3, characterized in that: Also included, in step S3, the specific steps include the following: Through the two-stage channel estimation process, the final channel estimation result is obtained, which is expressed as: Among them, λ is the estimated parameter adjusted according to the dynamic threshold.

9. The spatial non-stationary massive MIMO channel estimation method based on pre-screening and multi-level dynamic threshold strategy according to claim 3, characterized in that: The method further includes: simulation verification in step S4, verifying the performance of the estimation method through simulation analysis, and specific performance indicators include the number of atom selections, mean square error (NMSE) and reconstruction success probability.

Images