Mixed field channel estimation method for super-large-scale MIMO system

Abstract

The invention discloses a mixed field channel estimation method for a super-large-scale MIMO (Multiple Input Multiple Output) system. Combining the sparse transformation matrixes of the two channels into one sparse transformation matrix, and carrying out sparse representation on the mixed field channel; a mixed field channel estimation problem is converted into a sparse estimation problem, and the problem that two channels interfere with each other in the two sparse estimation processes is avoided. In the estimation process, the far-field sparse channel and the near-field sparse channel are jointly estimated according to the current support set every time the support set of the mixed field sparse channel is updated, and the residual error is calculated according to the current estimation value and used for updating the support set next time. The problem that mutual interference of a far-field sparse channel and a near-field sparse channel in the support set selection process cannot be solved in an existing scheme is solved. A simulation result shows that the scheme has higher estimation precision.

Description

Technical Field

[0001] The present invention belongs to the technical field of information and communication engineering, and relates to a very large-scale MIMO technology in a wireless communication system, and specifically to a mixed field channel estimation scheme based on compressed sensing theory in a very large-scale MIMO system. Background Art

[0002] As a potential key technology for the next generation of communication systems, the spectrum efficiency and communication rate of ultra-large-scale MIMO systems have been improved by dozens of times compared with large-scale MIMO. A high-precision channel estimation algorithm is a prerequisite for the performance of ultra-large-scale MIMO. In existing large-scale MIMO systems, since they are equipped with large-scale antenna arrays, the use of traditional least squares and minimum mean square error algorithms will bring huge pilot overhead and increase in computational complexity. The channel estimation scheme based on compressed sensing theory fills the gaps of the above schemes. The sparsity of the channel is used to achieve channel estimation with lower computational complexity and shorter pilot length than traditional schemes, effectively reducing the pilot overhead. Since the aperture of the antenna array at the base station end of the large-scale MIMO system is small, the near-field area is negligible. The communication between users and base stations is in the far-field area of ​​the base station. By utilizing the sparsity of the far-field channel in the angular domain, the sparse channel estimation based on compressed sensing theory takes into account both low overhead and high accuracy. Since XL-MIMO is equipped with a very large-scale antenna array at the base station, the number of antennas increases dramatically compared to large-scale MIMO, and the dimension of the channel also increases with the number of antennas. To achieve channel estimation, it will face higher computational complexity and pilot overhead. Therefore, in the very large-scale MIMO system, channel estimation based on compressed sensing theory has greater practical value. The near-field region of electromagnetic radiation of the very large-scale antenna array cannot be ignored, and channel estimation must take into account the near-field channel and its sparsity. In the actual communication environment, due to the scatterers distributed in the far-field region and the near-field region, the multipath channel contains both far-field channels and near-field channels. If the channel is regarded as a single type of channel for channel estimation, the accuracy of channel estimation is low, and the estimation result is not enough to reflect the channel situation. The existing mixed-field channel estimation method based on compressed sensing estimates the far-field channel and the near-field channel in the mixed-field channel separately, and there is a sequence of estimation. In the estimation process, there will be interference in the selection of support sets of far-field sparse channels and near-field sparse channels. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a mixed field channel estimation method for ultra-large-scale MIMO systems. The sparse transformation matrices of the two channels are merged into a sparse transformation matrix and the mixed field channel is sparsely represented; the mixed field channel estimation problem is converted into a sparse estimation problem, avoiding the problem of mutual interference between the two channels during the two sparse estimation processes. During the estimation process, each time the support set of the mixed field sparse channel is updated, the far-field sparse channel and the near-field sparse channel are jointly estimated based on the current support set, and the residual is calculated based on the current estimated value for the next update of the support set. It overcomes the problem that the existing scheme cannot solve the mutual interference between the far-field sparse channel and the near-field sparse channel during the support set selection process. Simulation results show that the scheme has higher estimation accuracy. The specific steps of the technical scheme adopted by the present invention to solve its technical problems are as follows:

[0004] A mixed field channel estimation method for a very large-scale MIMO system comprises the following steps:

[0005] Step 1: Determine the signal transmission model and mixed field channel model of the ultra-large-scale MIMO system and describe the problem;

[0006] The downlink communication process between users and base stations in the ultra-large-scale MIMO system is modeled. The near-field channel model, the far-field channel model, and the mixed-field channel model consisting of the two channels are determined.

[0007] Step 2: Establish a compressed sensing model for mixed field channel estimation.

[0008] The near-field sparse transformation matrix and the far-field sparse transformation matrix are combined so that the mixed-field channel can be sparsely represented as a single sparse signal, and the mixed-field channel estimation is transformed into a sparse signal estimation problem.

[0009] Step 3: Estimate the mixed field sparse channel based on the compressed sensing model of the mixed field channel estimation, and use the mixed field sparse channel to sparsely reconstruct the mixed field channel.

[0010] Furthermore, the signal transmission model of the ultra-large-scale MIMO system is as follows:

[0011] Consider the communication process between a base station equipped with N uniform linear antenna arrays and a single-antenna user in a very large-scale MIMO system, and perform channel estimation on its downlink. Assume that the base station sends a pilot symbol p in the mth time slot. m ∈C 1×N , then in the mth time slot, the pilot signal received by the user is in The mean is 0 and the variance is σ 2 The channel is represented by a column vector h with a dimension of N.m ∈C N×1 Denotes. Define M as the pilot length, m∈1,2,…M. Assuming that the channel remains unchanged during the pilot transmission phase, the signal received by the user end is:

[0012]

[0013] definition is the pilot matrix, the above formula can be further expressed as:

[0014] y=Ph+n (2)

[0015] where y=[y1,y2,…,y M ] H , is the signal received by the receiving end during the entire pilot transmission phase; n=[n1,n2,…,n M ] H , is the noise at the receiving end. The present invention estimates the mixed field channel h based on the known y and the pilot matrix P.

[0016] Furthermore, the near-field channel model, the far-field channel model and the mixed-field channel model composed of the two channels are determined as follows:

[0017] Line-of-sight channel between the nth antenna at the base station and the user [h] n Can be expressed as:

[0018]

[0019] Where α is the path attenuation, and it is assumed that the path attenuation is the same for all antennas; f c is the carrier frequency; r (n) is the distance between the nth antenna of the base station and the user; c is the speed of light. Therefore, the line-of-sight channel h between the user and the ULA with N antennas LOS Can be expressed as:

[0020]

[0021] In wireless communication systems, electromagnetic radiation is measured by the Rayleigh Distance It is divided into far field area and near field area, where D is the aperture of the antenna array and λ is the wavelength of the electromagnetic wave. When , the user is in the far field. The electromagnetic wave front is modeled by plane wave, and the arrival angle / departure angle is θ. Then the path difference of electromagnetic propagation between two adjacent antennas is dcosθ, and the path difference between the nth antenna and the first antenna is (n-1)dcosθ. (1) is the distance between the user and the reference antenna, far-field line-of-sight channel It can be further expressed as:

[0022]

[0023] in:

[0024]

[0025] Far-field multipath channel h far-field can be expressed as:

[0026]

[0027] Where L is the number of multipaths, α l is the gain of the lth path. The far-field channel is sparse in the angle domain and can be sparsely represented in the angle domain, namely:

[0028] h far-field =Fh a (8)

[0029] Where F is the discrete Fourier matrix, h a is the angle domain sparse channel.

[0030] When the distance between the user and the base station When the user is in the near field area, the plane wave model cannot accurately describe the near field channel, and the spherical wave model should be used. The distance between the user end and the first antenna of the array is r, the arrival angle / departure angle is θ, and the near field line-of-sight channel Can be expressed as:

[0031]

[0032] in is the array response vector of the near-field channel. The distance between the user and the nth antenna of the base station:

[0033]

[0034] Near-field multipath channel h near-field Can be expressed as:

[0035]

[0036] The near-field channel is sparse in the polar region, that is:

[0037] h near-field =Wh p (12)

[0038] where W∈C N×S is the polar domain transformation matrix, S nh is the number of samples of distance in different directions. p is the near-field multipath channel h near-field Sparse representation in the polar domain. W is composed of near-field array response vectors with different angle and distance parameters, as follows:

[0039]

[0040] In ultra-large-scale MIMO systems, considering the multipath effects caused by scatterers distributed in both the far field and the near field, the actual mixed-field multipath channel can be modeled as:

[0041]

[0042] Where L = L f +L n , L is the total number of multipaths in the mixed field channel, L f and L n represent the number of multipaths in the far-field channel and the near-field channel respectively. At the same time, the far-field path gain and near-field path gain of the lth path are and

[0043] Furthermore, the specific method of step 2 is as follows:

[0044] The compressed sensing model of mixed field channel estimation is established by equations (2), (8), (12), and (14):

[0045] y=PΨ h h+n (15)

[0046] Where Ψ=[W,F], which is the sparse transformation matrix of the mixed field channel. Therefore, Ψ is the sparse transformation matrix of the far field channel h f and the near-field channel h n The sparse representation of is sufficient. That is, h n =Ψh' p ,h f =Ψh' a , where h' a ,h' p ∈C (N+S)×1 ,h' a is the sparse representation of the far-field channel under the sparse transformation matrix of the mixed-field channel, h' p is the sparse representation of the near-field channel under the sparse transformation matrix of the hybrid channel,

[0047] Furthermore, the specific method of step 3 is as follows:

[0048] Under the compressed sensing model of formula (15), for the mixed field sparse channel h hThe sparse reconstruction is equivalent to the mixed field channel h. In the mixed field channel estimation, the column with the largest correlation between the residual and the perception matrix column is selected based on the orthogonal matching pursuit algorithm and added to the index set. The support set of the mixed field sparse channel is determined by the index set, and the non-zero coefficients of the near-field sparse channel and the non-zero coefficients of the far-field coefficient channel are jointly determined, and the residual is updated at the same time.

[0049] Furthermore, the specific mixed field channel estimation algorithm is as follows:

[0050] Step 1: Construct the mixed field sparse channel sensing matrix A = P[W, F] = PΨ; initialize the residual r (0) =y; Initialize index set Calculate the near-field channel perception matrix A n =PW; far-field channel sensing matrix A f =PF; Initialization loop number i=0.

[0051] Step 2: Calculate the residual r (i) The inner product of each column of the perception matrix A. And select the column corresponding to the column vector in A that makes the inner product the largest as the index, recorded as n * .

[0052] Step 3: Set index n * Add to the index set Λ.

[0053] Step 4: Classify the index set Λ: If the value of an element in the index set Λ is less than or equal to S, it is classified into the index set Λ n , if the value of an element in the index set Λ is greater than S, it is divided into the index set Λ f .

[0054] Step 5: Utilize A n and the index set Λ n Construct index matrix A n (:,Λ n ), by Λ n Select h p The support set sup(h p ), the support set coefficients are calculated using the least squares method

[0055] Step 6: Update intermediate residuals

[0056] Step 7: Utilize A f and the index set Λ f Construct index matrix A f (:,Λ f ), by Λ f Select ha The support set sup(h a ), the support set coefficients are calculated using the least squares method

[0057] Step 8: Update residuals

[0058] Step 9: Repeat steps 2 to 8 based on the updated residual until the set number of cycles L is reached, i.e., i=L-1.

[0059] Step 10: Using the last loop calculation and Sparse reconstruction of mixed field channels:

[0060]

[0061] In actual mixed-field communication scenarios, it is easier to obtain the total number of multipath channels than to obtain the number of multipath channels of far-field channels and near-field channels respectively. The solution proposed in the present invention avoids using the prior condition of the number of multipath channels of near-field channels and far-field channels, and instead uses the total number of multipath channels as the number of algorithm iterations to indirectly reconstruct the mixed-field channel h, which is more practical. h When the support set of W is equal to the sparse channel h in the extreme domain, p and the sparse channel h in the angular domain under F a The support set.

[0062] The beneficial effects of the present invention are as follows:

[0063] The present invention aims at the problem that the sparsity of sparse channels cannot be accurately determined due to two reconstructions in the existing methods, improves the sparse recovery model of mixed field channels, designs a mixed field channel joint estimation scheme, jointly estimates the far field sparse channels and near field sparse channels in the current iteration in each iteration, and updates the residual according to the current far field channel and near field channel estimation values, thereby improving the estimation accuracy without significantly increasing the complexity. At the same time, it avoids the need for prior knowledge of the number of far and near field paths of the mixed field channel, and has more practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is the far-field channel model of uniform linear array.

[0065] Figure 2 It is the near-field channel model under uniform linear array.

[0066] Figure 3 This is a flow chart of the mixed field channel estimation algorithm proposed by the present invention.

[0067] Figure 4 N = 512, M = 256, and the number of far-field multipaths is equal to the number of near-field multipaths (L f =L n =5), the normalized mean square error simulation diagram under different signal-to-noise ratios.

[0068] Figure 5 N = 256, M = 128, and the number of far-field multipaths is equal to the number of near-field multipaths (L f =L n =5), the normalized mean square error simulation diagram under different signal-to-noise ratios.

[0069] Figure 6 The normalized mean square error simulation diagram under different far-field and near-field multipath numbers when N=512, M=256, and signal-to-noise ratio SNR=5. f = kL, the number of near-field channel multipaths L n =(1-k)L.

[0070] Figure 7 The normalized mean square error simulation diagram under different far-field and near-field multipath numbers when N=512, M=128, and signal-to-noise ratio SNR=5. f = kL, the number of near-field channel multipaths L n =(1-k)L. DETAILED DESCRIPTION

[0071] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0072] A mixed field channel estimation method for a very large-scale MIMO system comprises the following steps:

[0073] Step 1: Determine the signal transmission model and mixed field channel model of the ultra-large-scale MIMO system and describe the problem;

[0074] The downlink communication process between users and base stations in a very large-scale MIMO system is modeled. Consider the communication process between a base station equipped with N uniform linear antenna arrays and a single-antenna user in a very large-scale MIMO system, and perform channel estimation on its downlink. Assume that the base station sends a pilot symbol p in the mth time slot. m ∈C 1×N , then in the mth time slot, the pilot signal received by the user is in The mean is 0 and the variance is σ 2 The channel is represented by a column vector h with a dimension of N. m ∈C N×1Denotes. Define M as the pilot length, m∈1,2,…M. Assuming that the channel remains unchanged during the pilot transmission phase, the signal received by the user end is:

[0075]

[0076] definition is the pilot matrix, the above formula can be further expressed as:

[0077] y=Ph+n (2)

[0078] where y=[y1,y2,…,y M ] H , is the signal received by the receiving end during the entire pilot transmission phase; n=[n1,n2,…,n M ] H , is the noise at the receiving end. The present invention estimates the mixed field channel h based on the known y and the pilot matrix P.

[0079] The near-field channel model, the far-field channel model and the mixed-field channel model consisting of the two channels are given below.

[0080] Line-of-sight channel between the nth antenna at the base station and the user [h] n Can be expressed as:

[0081]

[0082] Where α is the path attenuation, and it is assumed that the path attenuation is the same for all antennas; f c is the carrier frequency; r (n) is the distance between the nth antenna of the base station and the user; c is the speed of light. Therefore, the line-of-sight channel h between the user and the ULA with N antennas LOS Can be expressed as:

[0083]

[0084] In wireless communication systems, electromagnetic radiation is measured by the Rayleigh Distance It is divided into far field area and near field area, where D is the aperture of the antenna array and λ is the wavelength of the electromagnetic wave. , the user is in the far field area. Figure 1 For the far-field channel model of a uniform linear array, electromagnetic propagation is modeled as a plane wave that arrives at / leaves each antenna on the array at the same angle θ, where d is the antenna spacing. The electromagnetic wave front is modeled using a plane wave with an arrival / departure angle of θ. The path difference of electromagnetic propagation between two adjacent antennas is dcosθ, and the path difference between the nth antenna and the first antenna is (n-1)dcosθ. (1)is the distance between the user and the reference antenna, far-field line-of-sight channel It can be further expressed as:

[0085]

[0086] in:

[0087]

[0088] Far-field multipath channel h far-fileld can be expressed as:

[0089]

[0090] Where L is the number of multipaths, α l is the gain of the lth path. The far-field channel is sparse in the angle domain and can be sparsely represented in the angle domain, namely:

[0091] h far-field =Fh a (8)

[0092] Where F is the discrete Fourier matrix, h a is the angle domain sparse channel.

[0093] When the distance between the user and the base station , the user is in the near field area. Figure 2 is the near-field channel model for a uniform linear array. Electromagnetic propagation is modeled as a spherical wave that arrives at / leaves each antenna on the array at different angles. (N) is the distance from the user / scatterer to the Nth antenna in the array. The plane wave model cannot accurately describe the near-field channel, and spherical wave modeling should be used. The distance between the user end and the first antenna in the array is r, the arrival angle / departure angle is θ, and the near-field line-of-sight channel Can be expressed as:

[0094]

[0095] in is the array response vector of the near-field channel. The distance between the user and the nth antenna of the base station:

[0096]

[0097] Near-field multipath channel h near-field Can be expressed as:

[0098]

[0099] The near-field channel is sparse in the polar region, that is:

[0100] h near-field =Wh p (12)

[0101] where W∈C N×S is the polar domain transformation matrix, S n h is the number of samples of distance in different directions. p is the near-field multipath channel h near-field Sparse representation in the polar domain. W is composed of near-field array response vectors with different angle and distance parameters, as follows:

[0102]

[0103] In ultra-large-scale MIMO systems, considering the multipath effects caused by scatterers distributed in both the far field and the near field, the actual mixed-field multipath channel can be modeled as:

[0104]

[0105] Where L = L f +L n , L is the total number of multipaths in the mixed field channel, L f and L n represent the number of multipaths in the far-field channel and the near-field channel respectively. At the same time, the far-field path gain and near-field path gain of the lth path are and

[0106] Step 2: Establish a compressed sensing model for mixed field channel estimation.

[0107] By combining the near-field sparse transformation matrix and the far-field sparse transformation matrix, the mixed-field channel can be sparsely represented as a single sparse signal, and the mixed-field channel estimation is transformed into a sparse signal estimation problem. The details are as follows: The compressed sensing model of mixed-field channel estimation is established by equations (2), (8), (12), and (14):

[0108] y=PΨh h +n (15)

[0109] Where Ψ=[W,F], which is the sparse transformation matrix of the mixed field channel. Therefore, Ψ is the sparse transformation matrix of the far field channel h f and the near-field channel h n The sparse representation of is sufficient. That is, h n =Ψh' p ,hf=Ψh' a , where h' a ,h' p ∈C (N+S)×1 ,h' ais the sparse representation of the far-field channel under the sparse transformation matrix of the mixed-field channel, h' p is the sparse representation of the near-field channel under the sparse transformation matrix of the hybrid channel,

[0110] Step 3: Estimate the mixed field sparse channel based on the compressed sensing model of the mixed field channel estimation, and use the mixed field sparse channel to sparsely reconstruct the mixed field channel.

[0111] Under the compressed sensing model of formula (15), for the mixed field sparse channel h h The sparse reconstruction is equivalent to the mixed field channel h. In the mixed field channel estimation, the column with the largest correlation between the residual and the perception matrix column is selected based on the orthogonal matching pursuit algorithm and added to the index set. The support set of the mixed field sparse channel is determined by the index set, and in the process of calculating the coefficients, the non-zero coefficients of the near-field sparse channel and the non-zero coefficients of the far-field coefficient channel are jointly determined, and the residual is updated at the same time. Figure 3 As shown, the specific mixed field channel estimation algorithm is as follows:

[0112] Step 1: Construct the mixed field sparse channel sensing matrix A = P[W, F] = PΨ; initialize the residual r (0) =y; Initialize index set Calculate the near-field channel perception matrix A n =PW; far-field channel sensing matrix A f=PF; initialization cycle number i=0.

[0113] Step 2: Calculate the residual r (i) The inner product of each column of the perception matrix A. And select the column corresponding to the column vector in A that makes the inner product the largest as the index, recorded as n * .

[0114] Step 3: Set index n * Add to the index set Λ.

[0115] Step 4: Classify the index set Λ: If the value of an element in the index set Λ is less than or equal to S, it is classified into the index set Λ n , if the value of an element in the index set Λ is greater than S, it is divided into the index set Λ f .

[0116] Step 5: Utilize A n and the index set Λ n Construct index matrix A n (:,Λ n ), by Λ n Select h p The support set sup(h p), the support set coefficients are calculated using the least squares method

[0117] Step 6: Update intermediate residuals

[0118] Step 7: Utilize A f and the index set Λ f Construct index matrix A f (:,Λ f ), by Λ f Select h a The support set sup(h a ), the support set coefficients are calculated using the least squares method

[0119] Step 8: Update residuals

[0120] Step 9: Repeat steps 2 to 8 based on the updated residual until the set number of cycles L is reached, i.e., i=L-1.

[0121] Step 10: Using the last loop calculation and Sparse reconstruction of mixed field channels:

[0122]

[0123] In actual mixed-field communication scenarios, it is easier to obtain the total number of multipath channels than to obtain the number of multipath channels of far-field channels and near-field channels respectively. The solution proposed in the present invention avoids using the prior condition of the number of multipath channels of near-field channels and far-field channels, and instead uses the total number of multipath channels as the number of algorithm iterations to indirectly reconstruct the mixed-field channel h, which is more practical. h When the support set of W is equal to the sparse channel h in the extreme domain, p and the sparse channel h in the angular domain under F a The support set.

[0124] The complexity analysis of the solution proposed in the present invention is as follows: In the Lth iteration, step 2 involves N+S calculations of vector inner products, and its complexity is Steps 5 to 8 involve two pseudo-inverse calculations and matrix multiplications, and their complexity is Perform L iterations, the total complexity is Considering L n ,L f ,L is much smaller than M,N,S, and the computational complexity of this scheme is recorded as

[0125] Example

[0126] The present invention sets the required system parameters and algorithm initial values. The carrier frequency is 30 GHz, the number of multipath channels L is 10, and the path gain of the multipath channel obeys a Gaussian distribution with a mean of 0 and a variance of 1, that is, The arrival angle / departure angle of the multipath channel is uniformly distributed between 0 and π, that is, Near-field users / scatterers are evenly distributed 10 to 80 meters from the base station.

[0127] In order to verify the performance of the scheme proposed in the present invention under different signal-to-noise ratios, a mixed field channel is first generated according to the above parameters. Among them, the number of multipaths in the near-field channel is L n and the number of multipaths in the far-field channel L f All are 5, and a random pilot P is used. Add additive Gaussian white noise of different powers at the receiving end, and the signal received by the receiving end is y=Ph+n. Adjust the noise power, and use the received signal and the known pilot to estimate h when the signal-to-noise ratio SNR is 0, 1, 2...10dB, and record the estimated value Repeat the simulation 1000 times and calculate the normalized mean square error between the estimated value and the true value

[0128] Figure 4 N = 512, M = 256, and the number of far-field multipaths is equal to the number of near-field multipaths (L f =L n =5), the normalized mean square error simulation diagram under different signal-to-noise ratios. As the signal-to-noise ratio increases, the normalized mean square error of different estimation schemes gradually decreases. The normalized mean square error of the estimation scheme proposed in the present invention under different signal-to-noise ratios is better than that of the existing schemes. Referring to the minimum mean square error scheme, the signal-to-noise ratio is improved at a low signal-to-noise ratio, and the effect is more significant. When N = 256 and M = 128, Figure 5 As shown, the solution proposed by the present invention is still better than the existing solution.

[0129] In order to verify the performance of the scheme proposed in the present invention under different hybrid channel multipath channels, the noise power is fixed so that the signal-to-noise ratio SNR is 5, and the number of multipath channels of different types in the hybrid field channel h and the number of near-field multipath channels L are adjusted. n =(1-k)L, the number of far-field channel multipaths L f= kL. Each time a near-field path is reduced and a far-field multipath channel is added until the number of far-field channel multipaths is 10, that is, k = 0, 0.1, 0.2 ... 1. In each case, a channel h is generated, and a random pilot P is used. After the pilot is transmitted through the channel, the receiving end receives y = Ph + n. Using the received signal and the known pilot, h is estimated and the estimated value of the channel is recorded. Repeat the simulation 1000 times, calculate the normalized mean square error between the estimated value and the true value, and compare the normalized mean square error of different estimation schemes

[0130] Figure 6 The normalized mean square error of the far-field channel and the near-field channel under different multipath numbers when N=512, M=256, and signal-to-noise ratio SNR=5dB. The scheme proposed in the present invention has a lower normalized mean square error under different multipath numbers of far-field and near-field channels. This shows that the scheme proposed in the present invention has better performance in mixed-field channel estimation and is more adaptable to mixed-field multipath channels in different situations. The simulation under the parameters of N=256 and M=128 is as follows Figure 7 As shown, the estimation accuracy of the solution proposed by the present invention is still better than that of the existing solutions.

Claims

1. A mixed field channel estimation method for ultra-large-scale MIMO systems, characterized in that: The steps include: Step 1: Determine the signal transmission model and mixed field channel model of the ultra-large-scale MIMO system and describe the problem; Model the downlink communication process between users and base stations in ultra-large-scale MIMO systems; Determine a near-field channel model and a far-field channel model as well as a mixed-field channel model consisting of the two channels; Step 2: Establish a compressed sensing model for mixed field channel estimation; Combining the near-field sparse transformation matrix and the far-field sparse transformation matrix, the mixed-field channel can be sparsely represented as a single sparse signal, and the mixed-field channel estimation is transformed into a sparse signal estimation problem; Step 3: Estimate the mixed field sparse channel based on the compressed sensing model of the mixed field channel estimation, and use the mixed field sparse channel to sparsely reconstruct the mixed field channel.

2. A mixed field channel estimation method for a very large-scale MIMO system according to claim 1, characterized in that: The signal transmission model of the ultra-large-scale MIMO system is as follows: Consider the communication process between a base station equipped with N uniform linear antenna arrays and a single-antenna user in a very large-scale MIMO system, and perform channel estimation on its downlink. Assume that the base station sends a pilot symbol p in the mth time slot. m ∈C 1×N , then in the mth time slot, the pilot signal received by the user is in The mean is 0 and the variance is σ 2 The channel is represented by a column vector h with a dimension of N. m ∈C N×1 Indicates; define M as the pilot length, m∈1,2,…M, assuming that the channel remains unchanged during the pilot transmission phase, the signal received by the user end is: definition is the pilot matrix, the above formula can be further expressed as: y=Ph+n (2) where y=[y1,y2,…,y M ] H , is the signal received by the receiving end during the entire pilot transmission phase; n=[n1,n2,…,n M ] H , is the noise at the receiving end; based on the known y and pilot matrix P, estimate the mixed field channel h.

3. A mixed field channel estimation method for a very large-scale MIMO system according to claim 2, characterized in that: The near-field channel model, the far-field channel model, and the mixed-field channel model consisting of the two channels are determined as follows: Line-of-sight channel between the nth antenna at the base station and the user [h] n Can be expressed as: Where α is the path attenuation, and it is assumed that the path attenuation is the same for all antennas; f c is the carrier frequency; r (n) is the distance between the nth antenna of the base station and the user; c is the speed of light; therefore, the line-of-sight channel h between the user and the ULA of N antennas LOS Can be expressed as: In wireless communication systems, electromagnetic radiation is measured by the Rayleigh distance It is divided into far-field area and near-field area, where D is the aperture of the antenna array and λ is the wavelength of the electromagnetic wave. When , the user is in the far field area; the electromagnetic wave front is modeled by plane wave, and the arrival angle / departure angle is θ; then the path difference of electromagnetic propagation between two adjacent antennas is dcosθ, and the path difference between the nth antenna and the first antenna is (n-1)dcosθ; r (1) is the distance between the user and the reference antenna, far-field line-of-sight channel It can be further expressed as: in: Far-field multipath channel h far-field can be expressed as: Where L is the number of multipaths, α l is the gain of the lth path; the far-field channel is sparse in the angle domain, and it can be sparsely represented in the angle domain, that is: h far-field =Fh a (8) Where F is the discrete Fourier matrix, h a is a sparse channel in the angle domain; When the distance between the user and the base station When the user is in the near field area, the plane wave model cannot accurately describe the near field channel, and the spherical wave model should be used. The distance between the user end and the first antenna of the array is r, the arrival angle / departure angle is θ, and the near field line-of-sight channel Can be expressed as: in is the array response vector of the near-field channel; the distance between the user and the nth antenna of the base station: Near-field multipath channel h near-field Can be expressed as: The near-field channel is sparse in the polar region, that is: h near-field =Wh p (12) where W∈C N×S is the polar domain transformation matrix, S n is the number of samples of distance in different directions; h p is the near-field multipath channel h near-field Sparse representation in the polar domain; W is composed of near-field array response vectors with different angle and distance parameters, as follows: In ultra-large-scale MIMO systems, considering the multipath effects caused by scatterers distributed in both the far field and the near field, the actual mixed-field multipath channel can be modeled as: Where L = L f +L n , L is the total number of multipaths in the mixed field channel, L f and L n represent the number of multipaths in the far-field channel and the near-field channel respectively; meanwhile, the far-field path gain and near-field path gain of the lth path are and 4. A mixed field channel estimation method for ultra-large-scale MIMO system according to claim 3, characterized in that: Step 2: The compressed sensing model of mixed field channel estimation is established by equations (2), (8), (12), and (14): y=PΨh h +n (15) Where Ψ = [W, F], denoted as the sparse transformation matrix of the mixed field channel; therefore, Ψ is the sparse transformation matrix of the far field channel h f and the near-field channel h n The sparse representation of is sufficient; that is, h n =Ψh' p ,h f =Ψh' a , where h' a ,h' p ∈C (N+S)×1 ,h' a is the sparse representation of the far-field channel under the sparse transformation matrix of the mixed-field channel, h' p is the sparse representation of the near-field channel under the sparse transformation matrix of the hybrid channel, 5. A mixed field channel estimation method for ultra-large-scale MIMO system according to claim 4, characterized in that: Step 3: Under the compressed sensing model of formula (15), for the mixed field sparse channel h h The sparse reconstruction is equivalent to the mixed field channel h; in the mixed field channel estimation, the column with the largest correlation between the residual and the perception matrix column is selected based on the orthogonal matching pursuit algorithm and added to the index set; The support set of the mixed field sparse channel is determined by the index set and in the process of calculating the coefficients, the non-zero coefficients of the near-field sparse channel and the non-zero coefficients of the far-field coefficient channel are jointly determined, and the residual is updated at the same time.

6. A mixed field channel estimation method for a very large-scale MIMO system according to claim 5, characterized in that: The specific mixed field channel estimation algorithm is as follows: Step 1: Construct the mixed field sparse channel sensing matrix A = P[W, F] = PΨ; initialize the residual r (0) =y; Initialize index set Calculate the near-field channel perception matrix A n =PW; far-field channel sensing matrix A f =PF; Initialization cycle number i = 0; Step 2: Calculate the residual r (i) The inner product of each column of the perception matrix A; and select the column corresponding to the column vector in A that makes the inner product the largest as the index, denoted by n * ; Step 3: Set index n * Add to the index set Λ; Step 4: Classify the index set Λ: If the value of an element in the index set Λ is less than or equal to S, it is classified into the index set Λ n , if the value of an element in the index set Λ is greater than S, it is divided into the index set Λ f ; Step 5: Utilize A n and the index set Λ n Construct index matrix A n (:,Λ n ), by Λ n Select h p The support set sup(h p ), the support set coefficients are calculated using the least squares method Step 6: Update intermediate residuals Step 7: Utilize A f and the index set Λ f Construct index matrix A f (:,Λ f ), by Λ f Select h a The support set sup(h a ), the support set coefficients are calculated using the least squares method Step 8: Update residuals Step 9: Repeat steps 2 to 8 based on the updated residual until the set number of cycles L is reached, i.e., i=L-1; Step 10: Using the last loop calculation and Sparse reconstruction of mixed field channels:

Images