Channel estimation method and system combining Grassmann manifold in massive MIMO system

Abstract

The present invention discloses a channel estimation method and system in combination with Grassmann manifold in a large-scale MIMO system. The method of the present invention comprises the following steps: S1, a receiving end receives a signal sent by a base station, and obtains preliminary estimated channel information according to a preferred scheme based on the received signal; S2, constructs a merged Hankel matrix based on the preliminary estimated channel information and its conjugate data; S3, obtains an azimuth arrival angle estimation value using the properties of the Grassmann manifold; S4, restores channel state information using the azimuth arrival angle estimation value. The present invention utilizes a method for estimating channels using the properties of the Grassmann manifold. Compared with traditional methods, the method for estimating channels using the properties of the Grassmann manifold can reduce the influence of noise when a small amount of channel information is known, thereby achieving a higher channel utilization rate and further improving the performance of the communication system.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a channel estimation method and system combining Grassmann manifold in a large-scale MIMO system. Background Art

[0002] In a massive MIMO system, the performance of the wireless communication system is greatly affected by the wireless channel, such as shadow fading and frequency selective fading, which makes the propagation path between the transmitter and the receiver very complicated. The information transmitted by the wireless channel is regarded as a matrix. Using the concept of manifold, the transmitted channel information is a point in the Grassmann manifold. The wireless channel is not fixed and predictable like the wired channel, but has great randomness, so the signal sent from the base station to the user is also affected and becomes inaccurate. Therefore, how to obtain the complete CSI in the actual system is crucial to the system performance.

[0003] For single-user massive MIMO systems, obtaining downlink CSI requires a lot of pilot overhead. In actual MIMO systems, since the amount of training and feedback overhead is proportional to the number of BS antennas, the overhead of using traditional linear channel estimation methods (such as least squares (LS) and linear minimum mean square error (LMMSE)) to obtain CSI is too high.

[0004] In view of the above technical problems, it is necessary to improve them. Summary of the invention

[0005] In view of the above problems existing in the prior art, the present invention combines Grassmann manifold in the field of mathematics to provide a large-scale MIMO system channel estimation method and system, and the present invention improves the accuracy of channel estimation.

[0006] The present invention adopts the following technical solutions:

[0007] The channel estimation method combining Grassmann manifold in a massive MIMO system includes the following steps:

[0008] S1. The receiving end receives a signal sent by a base station, and obtains preliminary estimated channel information according to the received signal according to a preferred scheme;

[0009] S2, constructing a combined Hankel matrix based on the preliminary estimated channel information and its conjugate data;

[0010] S3, using the properties of the Grassmann manifold to obtain an azimuth arrival angle estimate;

[0011] S4. Restore channel state information using the azimuth arrival angle estimation value.

[0012] As a preferred solution, the channel information is preliminarily estimated in S1 The calculation formula is:

[0013]

[0014] Among them, L c represents the number of propagation paths from the base station to the user; β l represents the channel complex gain coefficient of the lth path; α(θ l ) represents the channel steering vector of the lth path; θ l is the azimuth arrival angle of the lth path; w is a zero-mean value with a variance of σ 2 Gaussian noise.

[0015] As a preferred solution, the channel steering vector of the lth path is expressed as:

[0016]

[0017] Among them, T represents transpose, N t represents the number of antennas installed at the base station, and j represents an imaginary unit.

[0018] As a preferred solution, in step S1, the base station sends the training sequence M times, and the receiving end obtains the preliminary estimated channel information based on the received signal as According to the above formula, it can be expressed as:

[0019]

[0020] As a preferred solution, step S2 includes the steps of:

[0021] S2.1, constructing a first Hankel matrix based on the preliminary estimated channel information;

[0022] S2.2, taking a conjugate transpose of the preliminary estimated channel information to obtain the preliminary estimated channel information after conjugate transposition;

[0023] S2.3, constructing a second Hankel matrix according to the preliminary estimated channel information after conjugate transposition;

[0024] S2.4. Merge the first Hankel matrix and the second Hankel matrix to obtain a merged Hankel matrix.

[0025] As a preferred solution, in step S2.1, the first Hankel matrix H QL It is expressed as:

[0026]

[0027] Obviously, H QL ∈C Q×L, C Q×L It is represented as the set of all complex matrices of size Q rows and L columns, where Q and L are positive integers and satisfy the conditions Q+L-1=M,Q≥L c ,L≥L c And Q ≥ L + 1, when M is an even number, then take When MH is an odd number, the an integer of ;

[0028] In step S2.2, the conjugate transpose of the preliminary estimated channel information is obtained It can be expressed as:

[0029]

[0030] The superscript H means taking the conjugate transpose;

[0031] In step S2.3, the second Hankel matrix It is expressed as:

[0032]

[0033] In step S2.4, the combined Hankel matrix is ​​expressed as:

[0034]

[0035] As a preferred solution, step S3 includes the following steps:

[0036] S3.1. Yes Perform SVD decomposition, that is Among them, U is a unitary matrix of size Q×Q, D is a diagonal matrix of size Q×2L, and V is a unitary matrix of size 2L×2L;

[0037] S3.2. Take out the first L of matrix U c (L c is the number of paths) column, and select 1 to 2L from the first row in sequence c The rows form the matrix T 1 , 2 to 2L c +1 row to form the matrix T 2 , 3 to 2L c +2 rows form the matrix T 3 Etc., and follow this rule until you get Q-2L c +1 to Q rows form a matrix

[0038] S3.3, respectively construct the matrix T constructed in step S3.2 k (k=1,2,…,Q-2L c +1) to the matrix The geodesic ψ k (t), where Is a size L c ×L c The zero matrix of Is a size L c ×L c The unit matrix of ψ k (t) is calculated as follows:

[0039]

[0040] Among them, U 1k , D 1k , V 1k is the matrix The result after SVD decomposition is Φ k = atan(D 1k ).

[0041] S3.4. Calculate the point to matrix on the geodesic line of step S3.3 Grassmann critical angle The critical angle calculation formula is as follows:

[0042]

[0043]

[0044]

[0045] Among them, U 2k , D 2k , V 2k Yes The matrix obtained by performing simplified SVD decomposition, D 2k (i 1 ,i 2 ) represents the matrix D 2k The i 1 Row i 2 Elements of a column.

[0046] S3.5. Find a point W on the geodesic constructed in step S3.3. 1k Make it with the matrix The critical angle satisfies the relation Here are the steps:

[0047]

[0048] Computational findings Therefore, ψ k (t) is simplified to:

[0049]

[0050] The steps to solve the critical angle are as follows:

[0051]

[0052]

[0053] Among them, Φ k (i 1 ,i 2 ) represents the matrix Φ k The i 1 Row i 2 The elements of the column will Substitute ψ k (t), we can get W 1k Repeat step S3.3 to construct W 1k To Matrix Repeat the above steps in S3.5 and find a point on the new geodesic that matches the matrix The critical angle satisfies the relation And so on, repeat the above steps until you find a point to the matrix The critical angle satisfies the relation And record this point as G 1 Similarly, according to the other geodesics in step S3.3, we can get matrix Replace it with other similar matrices such as matrix A new batch of estimated points can be obtained

[0054] S3.6. Let the total number of estimated points obtained in step S3.5 be g. Perform SVD decomposition on the g estimated points and extract the left singular vector matrix, which are recorded as P 1 ,P 2 ,…,P g .

[0055] S3.7. Calculate the matrix The eigenvalues ​​and corresponding eigenvectors of And there is Extract eigenvalues The corresponding eigenvectors form the matrix W.

[0056] S3.8, take 1 to 2L for W c -1 line is marked as U 1 , take 2 to 2L for W c The row is recorded as U 2 :

[0057] U1 =W(1:2L c -1,:)

[0058] U 2 =W(2:2L c ,:)

[0059] According to the formula:

[0060]

[0061] Among them, eig means taking the eigenvalue, and angle means the angle at which this eigenvalue is taken. The angle is recorded as but is θ i (i=1,2,…,L c ) is an estimated value.

[0062] As a preferred solution, step S4 includes:

[0063] S4.1. Based on the estimated value of the azimuth arrival angle Get the reconstructed channel steering vector

[0064] S4.2. Based on the reconstructed channel steering vector Estimate the corresponding reconstructed channel complex gain coefficient The specific steps are as follows:

[0065]

[0066]

[0067] S4.3. Restore channel state information based on the reconstructed channel steering vector and the reconstructed channel complex gain coefficient.

[0068] As a preferred solution, in step S4.3, the recovered channel state information The calculation formula is:

[0069]

[0070] in, That is the recovered channel state information,

[0071] The present invention also discloses a channel estimation system combining Grassmann manifold in a large-scale MIMO system, comprising the following modules:

[0072] Estimated channel information module: The receiving end receives the signal sent by the base station and obtains preliminary estimated channel information according to the received signal according to the preferred scheme;

[0073] Constructing Hankel matrix module: constructing a combined Hankel matrix based on the preliminary estimated channel information and its conjugate data;

[0074] Azimuth arrival angle estimation module: uses the properties of Grassmann manifold to obtain the azimuth arrival angle estimation value;

[0075] Channel state information recovery module: uses the azimuth arrival angle estimation value to restore the channel state information.

[0076] The beneficial effects of the present invention are:

[0077] Compared with the traditional method, the method for estimating channels using the properties of Grassmann manifolds in the present invention can reduce the influence of noise when a small amount of channel information is known, thereby achieving higher channel utilization and further improving the performance of the communication system. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0079] Figure 1 It is a simulation diagram of the average channel utilization of the method of the present invention under different signal-to-noise ratios (the number of training times is 33);

[0080] Figure 2 This is a simulation diagram of the average channel utilization of this method under different training times (the signal-to-noise ratio is 10dB);

[0081] Figure 3 This is a flow chart of a channel estimation method combining Grassmann manifold in a large-scale MIMO system in Example 1.

[0082] Figure 4 This is a block diagram of a channel estimation system combining Grassmann manifold in a large-scale MIMO system in Example 2. DETAILED DESCRIPTION

[0083] The following describes the implementation of the present invention through specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification.

[0084] Embodiment 1:

[0085] This embodiment illustrates the method of the present invention through a specific example. The present invention improves the channel estimation method for azimuth angle of arrival (AoA) estimation in a large-scale MIMO system.

[0086] The specific application cases are as follows:

[0087] Assume that there is one user, one base station and the number of antennas of the base station is 128. Table 1 below gives general parameter settings, and the AoA estimation and channel estimation are performed according to the parameters in Table 1.

[0088] Table 1 Parameter settings

[0089] parameter set up <![CDATA[Transmitting antenna N t > 128 <![CDATA[Receiving antenna N r > 1 The base station sends the training sequence times M 33 Signal-to-noise ratio 4,6,8,10,12,14,16,18 <![CDATA[Number of clusters L c > 4 Number of radiation paths p 1 <![CDATA[Matrix H QL Number of rows Q]]> 17 <![CDATA[Matrix H QL Number of columns L]]> 17

[0090] like Figure 3 As shown, the channel estimation method using channel conjugate data in the massive MIMO system of this embodiment includes the following steps:

[0091] S1. The receiving end receives a signal sent by a base station, and obtains preliminary estimated channel information according to the received signal according to a preferred scheme;

[0092] S2. constructing a combined Hankel matrix based on the preliminary estimated channel information and its conjugate data;

[0093] S3, splitting the merged Hankel matrix and combining it with the Grassmann manifold property to obtain the azimuth arrival angle estimation value;

[0094] S4. Restore channel state information using the azimuth arrival angle estimation value.

[0095] The present invention uses the property of Grassmann manifold to estimate the channel. Compared with the traditional method, when a small amount of channel information is known, the channel estimation method using conjugate data can reduce the influence of noise, thereby achieving higher channel utilization and further improving the performance of the communication system.

[0096] Specifically:

[0097] Assume that the massive antenna system includes one single-antenna user and one base station, and the base station is equipped with 128 antennas.

[0098] Step S1 is as follows: the channel preliminarily estimated by the user end based on the received signal is recorded as

[0099] In this large-scale antenna system, the base station antenna adopts a linear array (ULA) arrangement, and the channel from the base station to the user is expressed as:

[0100]

[0101] Among them, β l represents the channel complex gain coefficient of the lth path; α(θ l ) represents the channel steering vector of the lth path; θ l is the azimuth arrival angle of the lth path; w is a zero-mean value with a variance of σ 2 Gaussian noise.

[0102] The channel steering vector of the lth path is expressed as:

[0103]

[0104] Wherein, T represents transposition, j represents an imaginary unit, and in this embodiment j 2 =-1.

[0105] In step S1, the base station sends the training sequence 33 times, and the receiving end obtains the preliminary estimated channel information based on the received signal as It can be expressed as:

[0106]

[0107] Step S2 includes the following steps:

[0108] S2.1, constructing a first Hankel matrix based on the preliminary estimated channel information;

[0109] S2.2, taking a conjugate transpose of the preliminary estimated channel information to obtain the preliminary estimated channel information after conjugate transposition;

[0110] S2.3, constructing a second Hankel matrix according to the preliminary estimated channel information after conjugate transposition;

[0111] S2.4. Merge the first Hankel matrix and the second Hankel matrix to obtain a merged Hankel matrix.

[0112] In this embodiment, the first Hankel matrix H QL It is expressed as:

[0113]

[0114] H QL The matrix satisfies the conditions Q+L-1=32,Q≥4,L≥4, and Q is taken as 17.

[0115] In step S2.2, the conjugate transpose of the preliminary estimated channel information is obtained It is expressed as:

[0116]

[0117] The superscript H indicates taking the conjugate transpose.

[0118] It can be seen from this that and The corresponding generated column spaces are consistent.

[0119] In this embodiment, the second Hankel matrix in step S2.3 It is expressed as:

[0120]

[0121] In step S2.4, the combined Hankel matrix is ​​expressed as:

[0122]

[0123] Step S3 includes the steps of:

[0124] S3.1. Yes Perform SVD decomposition, that is Among them, U is a unitary matrix of size 17×17, D is a diagonal matrix of size 17×34, and V is a unitary matrix of size 34×34;

[0125] S3.2. Take out the first 4 columns of matrix U and select rows 1 to 8 from the first row to form matrix T 1 , 2 to 9 rows form the matrix T 2 , 3 to 10 rows form the matrix T 3 And so on, and follow this rule until 10 to 17 rows are obtained to form the matrix T 10 ;

[0126] S3.3, respectively construct the matrix T constructed in step S3.2 k (k=1,2,…,10) to matrix The geodesic ψ k (t), where 0 4×4 is a 4×4 zero matrix, I 4×4 is a unit matrix of size 4×4. k (t) is calculated as follows:

[0127]

[0128] Among them, U 1k , D 1k , V 1k is the matrix The result after SVD decomposition is

[0129] S3.4. Calculate the point to matrix on the geodesic line of step S3.3 Grassmann critical angle The details are as follows:

[0130]

[0131]

[0132]

[0133] Among them, U 2k , D 2k , V 2k Yes The matrix obtained by performing a reduced SVD decomposition.

[0134] S3.5. Substitute the geodesic ψ constructed in step S3.3 into k (t) can determine a point and record this point as W 1k , respectively construct W 1k To Matrix The geodesic ψ' k (t), ψ' k (t) is expressed as follows:

[0135]

[0136] Among them, U' 1k , D' 1k , V' 1k is the matrix The result after SVD decomposition is Φ' k = atan(D' 1k ).Will Substitute ψ' k (t) get a point denoted as G 1 Similarly, according to the other geodesics in step S3.3, we can get G by following the above steps. 2 ,G 3 ,…,G 10 .matrix Replace it with other similar matrices such as matrix A new batch of estimated points G' can be obtained 1 ,G' 2 ,…,G' 10 .

[0137] S3.6. Perform SVD decomposition on all 20 estimated points obtained in step S3.5, and extract the left singular vector matrix, which are denoted as P 1 ,P 2 ,…,P 20 .

[0138] S3.7. Calculate the matrix The eigenvalue and corresponding eigenvector of 1 ,λ 2 ,…,λ 8 , and there is λ 1 ≥λ 2 ≥…≥λ 8 . Take out the eigenvalue λ 1 ,λ 2 ,λ 3 ,λ 4 The corresponding eigenvectors form the matrix W.

[0139] S3.8. Take rows 1 to 7 of W and record them as U 1 , take rows 2 to 8 of W and record them as U 2 :

[0140] U 1 =W(1:7,:)

[0141] U 2 =W(2:8,:)

[0142] According to the formula:

[0143]

[0144] Among them, eig means taking the eigenvalue, and angle means the angle at which this eigenvalue is taken. The angle is recorded as but is θ i The estimated value of (i=1,2,3,4).

[0145] Step S4 comprises the steps of:

[0146] S4.1. Based on the estimated value of the azimuth arrival angle Get the reconstructed channel steering vector

[0147] S4.2. Based on the reconstructed channel steering vector Estimate the corresponding reconstructed channel complex gain coefficient Here are the steps:

[0148]

[0149]

[0150] S4.3. Restore channel state information based on the reconstructed channel steering vector and the reconstructed channel complex gain coefficient.

[0151] In step S4.3, the recovered channel state information The calculation formula is:

[0152]

[0153] in, That is, the restored channel, and the reconstructed channel steering vector of the lth path is Then, After normalization, according to the formula

[0154]

[0155] Obtain the recovered channel state information The utilization rate η.

[0156] Refer to the attached Figure 1 As shown, when the number of training times is 33 and the signal-to-noise ratio is 10 dB, the simulation of the channel estimation method of the present invention shows that the channel utilization rate is 81.9%, and as the system signal-to-noise ratio increases, the channel utilization rate also increases.

[0157] Refer to the attached Figure 2 As shown, when the signal-to-noise ratio is 10 dB, as the number increases, the channel utilization rate also increases and approaches 1.

[0158] Example 2

[0159] like Figure 4 As shown, this embodiment, based on the method of Embodiment 1, discloses a channel estimation system combining Grassmann manifold in a large-scale MIMO system, which includes the following modules:

[0160] Estimated channel information module: The receiving end receives the signal sent by the base station and obtains preliminary estimated channel information according to the received signal according to the preferred scheme;

[0161] Constructing Hankel matrix module: constructing a combined Hankel matrix based on the preliminary estimated channel information and its conjugate data;

[0162] Azimuth arrival angle estimation module: uses the properties of Grassmann manifold to obtain the azimuth arrival angle estimation value;

[0163] Channel state information recovery module: uses the azimuth arrival angle estimation value to restore the channel state information.

[0164] For other contents of this embodiment, please refer to Embodiment 1.

[0165] The present invention may also be implemented or applied through other different specific implementation methods, and the details in this specification may also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

Claims

1. Channel estimation method combining Grassmann manifold in massive MIMO system, Its characteristics are The steps include: S1. The receiving end receives the signal sent by the base station and obtains preliminary estimated channel information based on the received signal; S2, constructing a combined Hankel matrix based on the preliminary estimated channel information and its conjugate data; S3, using the properties of Grassmann manifold, obtain the azimuth arrival angle estimate; S4. Recover channel state information using the azimuth arrival angle estimation value; Step S2 specifically includes: S2.1, constructing a first Hankel matrix based on the preliminary estimated channel information; S2.2, taking a conjugate transpose of the preliminary estimated channel information to obtain the preliminary estimated channel information after conjugate transposition; S2.3, constructing a second Hankel matrix according to the preliminary estimated channel information after conjugate transposition; S2.4, merging the first Hankel matrix and the second Hankel matrix to obtain a merged Hankel matrix; In step S2.1, the first Hankel matrix H QL It is expressed as: Among them, H QL ∈C Q×L , C Q×L It is represented as the set of all complex matrices of size Q rows and L columns, where Q and L are positive integers and satisfy the conditions Q+L-1=M,Q≥L c ,L≥L c And Q ≥ L + 1, when M is an even number, then take An integer of, when M is an odd number, then take An integer of ; In step S2.2, the channel information is preliminarily estimated Taking the conjugate transpose, it is expressed as: The superscript H means taking the conjugate transpose; In step S2.3, the second Hankel matrix It is expressed as: In step S2.4, the combined Hankel matrix is ​​expressed as: Step S3 specifically includes: S3.

1. Yes Perform SVD decomposition, that is Among them, U is a unitary matrix of size Q×Q, D is a diagonal matrix of size Q×2L, and V is a unitary matrix of size 2L×2L; S3.

2. Take out the first L of matrix U c Column, L c is the number of paths, and select 1 to 2L from the first row in sequence c The rows form the matrix T 1 , 2 to 2L c +1 row to form the matrix T 2 , 3 to 2L c +2 rows form the matrix T 3 Etc., and follow this rule until you get Q-2L c +1 to Q rows form a matrix S3.3, respectively construct the matrix T constructed in step S3.2 k To Matrix The geodesic ψ k (t), k=1,2,…,Q-2L c +1, among which, Is a size L c ×L c The zero matrix of Is a size L c ×L c The unit matrix of k (t) is calculated as follows: Among them, U 1k , D 1k , V 1k is the matrix The result after SVD decomposition is Φ k = atan(D 1k ); S3.4, calculate the point on the geodesic of step S3.3 to the matrix Grassmann critical angle The critical angle calculation formula is as follows: Among them, U 2k , D 2k , V 2k Yes The matrix obtained by performing simplified SVD decomposition, D 2k (i 1 ,i 2 ) represents the matrix D 2k The i 1 Row i 2 Elements of a column; S3.

5. Find a point W on the geodesic constructed in step S3.

3. 1k Make it with the matrix The critical angle satisfies the relation Here are the steps: Computational findings Therefore, ψ k (t) is simplified to: The steps to solve the critical angle are as follows: Among them, Φ k (i 1 ,i 2 ) represents the matrix Φ k The i 1 Row i 2 The elements of the column will Substitute ψ k (t), we get W 1k ; Repeat step S3.3 to construct W 1k To Matrix Repeat step S3.5 and find a point on the new geodesic that matches the matrix The critical angle satisfies the relation And so on, repeat the above steps until you find a point to the matrix The critical angle satisfies the relation And record this point as G 1 Similarly, according to the other geodesics in step S3.3, we get matrix Replace it with other similar matrices such as matrix Get a new batch of estimated points S3.6, let the total number of all estimated points obtained in step S3.5 be g; perform SVD decomposition on the g estimated points respectively, and take out the left singular vector matrix, which is recorded as P 1 ,P 2 ,…,P g ; S3.

7. Calculate the matrix The eigenvalues ​​and corresponding eigenvectors of And there is Extract eigenvalues The corresponding eigenvectors form the matrix W; S3.8, take 1 to 2L for W c -1 line is marked as U 1 , take 2 to 2L for W c The row is recorded as U 2 : U 1 =W(1:2L c -1,:) U 2 =W(2:2L c ,:) According to the following formula: Among them, eig means taking the eigenvalue, and angle means the angle at which this eigenvalue is taken. The angle is recorded as but is θ i Estimated value of , i=1,2,…,L c ; In step S1, the channel information is preliminarily estimated The calculation formula is: Among them, L c represents the number of propagation paths from the base station to the user; β l represents the channel complex gain coefficient of the lth path; α(θ l ) represents the channel steering vector of the lth path; θ l is the azimuth arrival angle of the lth path; w is a zero-mean value with a variance of σ 2 Gaussian noise.

2. The channel estimation method in combination with Grassmann manifold in a massive MIMO system as claimed in claim 1, Its characteristics are: The channel steering vector of the lth path is expressed as: Among them, T represents transpose, N t represents the number of antennas installed at the base station, and j represents an imaginary unit.

3. The channel estimation method combining Grassmann manifold in a massive MIMO system as claimed in claim 2, Its characteristics are: In step S1, the base station sends the training sequence M times, and the receiving end obtains the preliminary estimated channel information based on the received signal as It is expressed as:

4. The channel estimation method in combination with Grassmann manifold in a massive MIMO system as claimed in claim 1, Its characteristics are: Step S4 specifically includes: S4.

1. Based on the estimated value of the azimuth arrival angle Get the reconstructed channel steering vector S4.

2. Based on the reconstructed channel steering vector Estimate the corresponding reconstructed channel complex gain coefficient The details are as follows: S4.

3. Restore channel state information based on the reconstructed channel steering vector and the reconstructed channel complex gain coefficient.

5. The channel estimation method in combination with Grassmann manifold in a massive MIMO system as claimed in claim 4, Its characteristics are: In step S4.3, the recovered channel state information The calculation formula is: in, is the recovered channel state information, 6. A channel estimation system incorporating Grassmann manifold in a massive MIMO system, wherein the channel estimation system is based on the method of claim 1, Its characteristics are: Includes the following modules: Estimated channel information module: The receiving end receives the signal sent by the base station and obtains preliminary estimated channel information based on the received signal; Constructing Hankel matrix module: constructing a combined Hankel matrix based on the preliminary estimated channel information and its conjugate data; Azimuth arrival angle estimation module: uses the properties of Grassmann manifold to obtain the azimuth arrival angle estimation value; Channel state information recovery module: uses the azimuth arrival angle estimation value to restore the channel state information.

Images