Super-large-scale MIMO uplink signal detection method suitable for near-field spherical waves

Abstract

The invention relates to the field of wireless communication, in particular to an ultra-large-scale MIMO uplink signal detection method suitable for near-field spherical waves, and in a near-field spherical wave model system, aiming at the problem of relatively large user interference caused by a communication scene with dense users, uplink signal detection specifically comprises the following steps: dividing a receiving antenna array into C non-intersecting sub-arrays, constructing a factor graph model by taking each array as a factor node and taking a sending signal of each user as a variable node; according to the factor graph model, approximate posterior distribution of a signal sent by a certain user on a subarray is obtained through calculation; and iteratively updating the received signal and the channel matrix of each sub-array based on a sorting interference elimination algorithm to complete signal detection. According to the method, better performance can be obtained in a multi-user near-field scene, and better detection performance can still be kept in a scene in which multiple users reach the center of the array at the same angle but at different distances.

Description

Technical Field

[0001] The present invention relates to the field of wireless communications, and in particular to a method for detecting near-field spherical wave ultra-large-scale MIMO uplink signals. Background Art

[0002] With the continuous development of mobile communication technology, the fifth generation mobile communication system (5G) has been deployed and applied worldwide. 5G has the characteristics of low latency, high speed, and strong reliability. However, the current performance is far from the ideal index. Therefore, researchers have started to study the sixth generation mobile communication system (6G).

[0003] Extremely large-scale Multiple-Input Multiple-Output (XL-MIMO) can improve the data detection rate by using spatial multiplexing gain, and improve the detection reliability by using spatial diversity gain, so that the system can serve multiple users at the same time and frequency resources. It has been considered as one of the key technologies of 6G wireless communication. The new characteristics of the ultra-large-scale MIMO channel are spatial non-stationarity and near-field characteristics. The spatial non-stationarity of the channel means that different areas of the array receive different powers of signals from different paths, and this difference is called the different degree of "visibility". Existing literature uses the molecular array method to study signal detection that adapts to non-stationarity. Because the ultra-large-scale MIMO system uses hundreds or even thousands of antennas to increase the size of the channel, the Rayleigh distance increases, and the user is in the near-field area, resulting in the difference in the angle of incidence of the user reaching each antenna unit that cannot be ignored. Therefore, channel modeling needs to consider a more general spherical wavefront model.

[0004] In the channel modeling of the plane wavefront hypothesis and the spherical wavefront hypothesis, both the distance parameter and the angle parameter play an important role, but there are still some differences. In the system using plane wavefront modeling, the distance parameter only affects the amplitude value, and when multiple users are at the same angle but different distances, there is a high correlation between the channels. This means that the channel characteristics between different users are similar, which increases the difficulty of signal detection and affects the system performance. In contrast, in the system of spherical wavefront modeling, the distance parameter affects both the amplitude value and the phase value. When multiple users are at the same angle but different distances, the channel correlation coefficient decreases as the distance difference between users increases. Therefore, this feature can be used to solve the signal detection problem in user-intensive scenarios and improve the detection performance. However, it can also be known that the channel correlation coefficient is still large for user-intensive scenarios, especially in scenarios where the angle between users is the same but the distance difference is small. In order to solve the problem of poor signal detection performance caused by the high correlation coefficient in user-intensive scenarios, a suitable signal detection algorithm is needed to improve the detection performance. Summary of the invention

[0005] Aiming at the signal detection problem in user-dense scenarios, the present invention proposes a method for detecting uplink signals of near-field spherical waves in ultra-large-scale MIMO. In a spherical wave model system, for user-dense communication scenarios, uplink signal detection specifically includes the following steps:

[0006] The receiving antenna array is divided into C non-overlapping sub-arrays, and each array is used as a factor node and each user's transmitted signal is used as a variable node to construct a factor graph model;

[0007] According to the factor graph model, the approximate posterior distribution of a signal sent by a certain user on a subarray is calculated;

[0008] Based on the sorting interference elimination algorithm, each subarray receiving signal and channel matrix are iteratively updated to complete signal detection.

[0009] Furthermore, in the factor graph model, the process of information transfer between factor nodes and variable nodes is expressed as:

[0010]

[0011] in, represents the symbol x sent by the jth user j The corresponding variable node to the cth subarray f c The information transmitted by the corresponding factor node; represents the symbol x sent by the j'th user j' The corresponding variable node to the cth subarray f c The information transmitted by the corresponding factor node, j'∈{1,2,...,K}\j, K is the number of users; Represents the symbol x sent from the variable node f0 corresponding to the prior information to the jth user j The corresponding variable node message; Represents the cth subarray f c The symbol x sent by the corresponding factor node to the jth user j The information transmitted by the corresponding variable node; Represents the c′th subarray f c′ The symbol x sent by the corresponding factor node to the jth user j The information transmitted by the corresponding variable node, c′∈{1,2,...,C}\c, {1,2,...,C}\c represents the set of {1,2,...,C} except c; x\x j represents the symbol set x sent by all users except the symbol x sent by the jth user j The symbol combination that constitutes c (xj ) represents the user x in the probability model j The marginal likelihood function of .

[0012] Furthermore, the approximate posterior distribution of the jth detected symbol in the cth subarray is It is expressed as:

[0013] Among them, b(x j ) represents the true posterior distribution of the jth detection symbol; represents the exponential family distribution; KL[A‖B] is the KL divergence operation, which is used to characterize the degree of deviation between distribution A and distribution B; represents the symbol x sent by the jth user j The corresponding variable node to the cth subarray f c The information transmitted by the corresponding factor node; Represents the cth subarray f c The symbol x sent by the corresponding factor node to the jth user j The information transmitted by the corresponding variable node.

[0014] Furthermore, the approximate posterior distribution b of all detected symbols in the cth subarray is fc The mean and variance of (x) are expressed as:

[0015]

[0016] Among them, μ c represents the approximate posterior distribution of all detected symbols received by the cth subarray The mean of c represents the approximate posterior distribution of all detected symbols in the cth subarray The variance of ;σ represents the noise variance; Denotes the c-th subarray channel matrix H c The transposed matrix of y c represents the received signal of the cth subarray; τ c represents the inverse of the prior variance of all detected symbols received by the cth subarray, γ c represents the a priori mean of all detected symbols received by the c-th subarray.

[0017] Furthermore, based on the sorting interference elimination algorithm, iteratively updating each subarray receiving signal and the channel matrix to eliminate the interference between users includes the following steps:

[0018] 101. Initialize the number of iterations to l = 0 and calculate the approximate posterior distribution b of all detected symbols in the cth subarray fc (x) the mean and variance;

[0019] 102. Let l=l+1, calculate the interference-to-signal ratio of each user in each subarray in the lth iteration, and arrange the detection order of users in each subarray according to the ascending sequence of the value to obtain the detection order of the cth subarray And let i=1;

[0020] 103. Select the test order The i-th user in is taken as the user to be estimated, and the received signal of the c-th subarray and the channel matrix of the c-th subarray are reconstructed using the mean of the approximate posterior distribution of all detected symbols of the i-th user in the c-th subarray;

[0021] 104. Determine whether i+1 is greater than K. If so, execute step 105. Otherwise, set i=i+1 and return to step 103.

[0022] 105. Determine whether the number of iterations l reaches the set maximum number of iterations or the algorithm converges, if so, end; otherwise, wait for the central processing unit to return data and repeat steps 102 to 104.

[0023] Further, reconstructing the received signal of the c-th subarray and the channel matrix of the c-th subarray by using the mean of the approximate posterior distribution of all detected symbols of the i-th user in the c-th subarray includes:

[0024]

[0025] in, Indicates that the signal of the i-th user in the detection sequence is calculated The received signal of the c-th subarray obtained by post-reconstruction is equal to the received signal of the c-th subarray when i=1, that is, represents the detection order in the channel matrix of the cth subarray at the i-th iteration The element corresponding to the i-th user in ; Indicates the detection order of the cth subarray in the i-th iteration The mean of the approximate posterior distribution corresponding to the i-th user in .

[0026] Furthermore, the calculation of the interference-to-signal ratio of each user in each subarray includes:

[0027]

[0028] Among them, JSR c,k represents the interference-to-signal ratio of the kth user in the cth subarray; h c,i represents the i-th column of the c-th subarray channel matrix, ||·|| represents the Euclidean norm, and K represents the number of users.

[0029] The present invention considers a base station equipped with a very large-scale array serving multiple single-antenna users, and considers the channel modeling method of spherical wavefront, thereby introducing a distance parameter; on the one hand, the present invention divides the channel into multiple sub-arrays and constructs a factor graph by utilizing the channel characteristics brought by the distance parameter; on the other hand, the continuous interference elimination algorithm is combined and adjusted with the expected propagation algorithm, and the iterative solution is finally obtained to obtain the detection symbol for demodulation. The present invention can achieve better performance in multi-user near-field scenarios, and can still maintain better detection performance in scenarios where multiple users arrive at the center of the array at the same angle but at different distances. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a schematic diagram of a multi-user near-field scenario of the present invention;

[0031] Figure 2 Schematic diagram of the factor graph model of the present invention;

[0032] Figure 3 A schematic diagram of a method for detecting a near-field spherical wave ultra-large-scale MIMO uplink signal according to the present invention;

[0033] Figure 4 This is a simulation diagram of the present invention using the bit error rate as a performance indicator of signal detection. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] In the spherical wave model system, when the users in the communication scene are dense, especially when the users are at the same angle and close to each other, the channel correlation is large, that is, the similarity between the channels is high. For this scenario, the present invention proposes a near-field spherical wave ultra-large-scale MIMO uplink signal detection method, such as Figure 3 , specifically including the following steps:

[0036] The receiving antenna array is divided into C non-overlapping sub-arrays, and each array is used as a factor node and each user's transmitted signal is used as a variable node to construct a factor graph model;

[0037] According to the factor graph model, the approximate posterior distribution of a signal sent by a certain user on a subarray is calculated;

[0038] Based on the sorting interference elimination algorithm, each subarray receiving signal and channel matrix are iteratively updated to complete signal detection.

[0039] like Figure 1 As shown, this embodiment considers that the transmitting end is K single-antenna users, and the receiving end is a large-scale uniform linear array base station equipped with N antennas to serve the users. For this scenario, based on the known channel state information and received signals, the sorted continuous interference elimination algorithm is combined with the distributed expectation propagation algorithm to complete the uplink signal detection task.

[0040] In this embodiment, the system modulation mode is χ-QAM, and the transmission symbol of the kth user is represented as x k ,x=[x1,x2,...,x K ] T ∈χ K is the symbol vector transmitted by K users in the same time slot. In this scenario, the uplink signal detection task is completed. The input and output signal model is:

[0041] y=Hx+n

[0042] in, is the received signal vector, is the channel matrix, is additive Gaussian white noise with mean 0 and variance σ 2 I N Distribution Complex Gaussian distribution, I N Represents the N×N identity matrix.

[0043] Under the spherical wave modeling, the near-field channel from the kth user to the receiving end is expressed as:

[0044]

[0045] Among them, b(θ k ,r k ) is the steering vector, r k represents the distance from the kth user to the center of the array in the LoS path, θ k represents the angle from the kth user to the center of the array in the LoS path, k c =2π / λ c Expressed as wave number, λ c =c / f c represents the wavelength. The steering vector assumed for the spherical wavefront is:

[0046]

[0047] in, represents the distance from user k to the nth antenna, β k represents the path loss of user k.

[0048] This embodiment considers a near-field channel scenario in which two or more users arrive at the center of the array at the same angle but at different distances. Figure 1 As shown. n =(2n-N+1) / 2, n=0,1,...,N-1, antenna spacing d, the distance from the user to the nth antenna is:

[0049]

[0050] in,

[0051] Consider that the N antennas at the receiving end are divided into C non-intersecting sub-array processing units, and the number of antennas in each sub-array is N c ,satisfy Considering the non-stationarity of the ultra-large-scale MIMO channel, each subarray unit is only visible to some users. The input and output signal model of the cth subarray is:

[0052] y c =H c x+n c

[0053] Among them, y c represents the signal vector received by the cth antenna subarray; represents an independent and identically distributed Rayleigh fading channel matrix with certain sparse columns; n c is the additive Gaussian white noise in the cth antenna subarray, which has a mean of 0 and a variance of The complex Gaussian distribution of N c ×N c The identity matrix of .

[0054] This embodiment only considers the signal detection part, and all subsequent work is performed under the condition that the receiving vector and the channel matrix are known. Under the condition that the receiving vector and the channel matrix are known, the posterior probability of the transmitted signal x is:

[0055]

[0056] in, represents the overall array received signal; p(y|x) and p(x) represent the likelihood function and prior distribution of x respectively; p(y c |x) represents the marginal likelihood function on the submatrix c; p(y) represents the prior information of the received signal y.

[0057] Consider combining the factor graph with the posterior probability, so the posterior probability of the transmitted signal x can be expressed as:

[0058]

[0059] Among them, f0(x) represents the prior distribution p(x); represents the likelihood function p(y|x); f c (x) can be equivalent to a marginal likelihood p(y c |x);

[0060] The factor graph model corresponding to this probability model is as follows Figure 2 As shown in the figure, the square represents the factor node (FN) and the circle represents the variable node (VN). According to the sum-product algorithm, the message passing process between FN and VN can be expressed as:

[0061]

[0062]

[0063] Among them, the symbol Represents the information transmitted from VN to FN; symbol It represents the information transmitted from FN to VN.

[0064] By definition, the marginal probability distribution of a variable node is equal to the product of all the input messages arriving at that node. Therefore, the transmitted signal x of the jth user in the cth subarray is j The actual posterior probability is:

[0065]

[0066] The above equation can be converted to:

[0067]

[0068] Defining confidence It belongs to the exponential family distribution It is used to represent the approximate posterior distribution of the jth detected symbol in the cth subarray, which is given by:

[0069]

[0070] Here, KL[·||·] represents the KL divergence operation, which is used to characterize the degree of deviation between two distributions. Minimizing the KL divergence of two distributions is equivalent to moment matching between the two distributions. Therefore, the distribution approximation process can be completed by matching the first-order moment and the second-order moment between the two distributions, thereby obtaining an approximate posterior distribution.

[0071] Approximate posterior distribution The mean and variance Σ c It can be generated by Bayesian MMSE estimation, that is:

[0072]

[0073] Among them, τ c =τ c I k and γ c =[γ 1c ,γ 2c ,…,γ Kc ] T denote the inverse of the prior variance and the prior mean of x on the cth submatrix, respectively. Assume that each approximate marginal distribution b(x j ) have the same variance, so Represents the average variance of the approximate marginal distribution, that is, ω c represents the approximate marginal distribution b(x j ), the inverse of the mean square error, ω c Take Σ c The reciprocal of the average diagonal element of c =((1 / K)tr(Σ c )) -1 . The mean value is μ c The element μ corresponding to the index c,j ,so and The prior confidence provided by the factor node f0 is defined as ω0 and μ0 are the symbol variance and symbol mean fused by the central processing unit and will be fed back to other factor nodes as prior information. After completing the moment matching, we can get:

[0074]

[0075] in, Right now The mean is μ c,j , the variance is (v c,j ) -1 Gaussian distribution μ c,j Represents factor node f c Passed to variable node x j The mean value of (v c,j ) -1 Represents factor node f c Passed to variable node x j The variance of Right now The mean is μ j,c , the variance is (v j,c ) -1 Gaussian distribution μj,c Represents variable node x j Pass back to factor node f c The prior mean of (v j,c ) -1 Represents variable node x j Pass back to factor node f c The prior variance of the factor node f0 can be calculated in the same way to obtain the information transmitted back to the variable node. The factor graph model used in the present invention is as follows: Figure 2 As shown, the CPU end is used as a factor node, each sending end is used as a variable node, and each subarray is used as a factor node. The transfer relationship between the variable node and the factor node is obtained through the factor graph model.

[0076] This embodiment considers that in a multi-user near-field scenario, the channel correlation of spherical wavefront modeling is related to the distance parameter. As the distance difference between users to the center of the array gradually increases, the correlation shows a downward trend and approaches 0. But it also illustrates that when the distance difference between users is small, it will still be affected by large signal interference. In order to eliminate interference between users, the ordered successive interference cancellation algorithm (OSIC) is combined with the distributed EP algorithm to obtain the EP-OSIC detection algorithm based on the factor graph. It mainly combines the MMSE estimation part with OSIC. Therefore, the estimated mean of each subarray is updated as follows:

[0077] 101. Initialize the number of iterations to l = 0 and calculate the approximate posterior distribution of all detected symbols in the cth subarray The mean and variance of Update the mean of the confidence;

[0078] 102. Let l=l+1, calculate the interference-to-signal ratio of each user in each subarray in the lth iteration, and arrange the detection order of users in each subarray according to the ascending sequence of the value to obtain the detection order of the cth subarray And let i=1; specifically:

[0079] Calculate the interference-to-signal ratio of each user in each subarray, including:

[0080]

[0081] Among them, JSR c,k represents the interference-to-signal ratio of the kth user in the cth subarray; h c,i represents the i-th column of the c-th subarray channel matrix, ||·|| represents the Euclidean norm;

[0082] 103. Select the test order The i-th user in is taken as the user to be estimated, and the received signal of the c-th subarray and the channel matrix of the c-th subarray are reconstructed using the mean of the approximate posterior distribution of all detected symbols of the i-th user in the c-th subarray; specifically:

[0083] Calculate the detection order The mean μ of the approximate posterior distribution of all detected symbols of the i-th user in the c-th subarray is c,i , use this signal to reconstruct the channel matrix and received signal of the cth subarray, that is:

[0084]

[0085] in, Indicates that the signal of the i-th user in the detection sequence is calculated The received signal of the c-th subarray obtained by post-reconstruction is equal to the received signal of the c-th subarray when i=1, that is, represents the detection order in the channel matrix of the cth subarray at the i-th iteration The element corresponding to the i-th user in ; Indicates the detection order of the cth subarray in the i-th iteration The mean of the approximate posterior distribution corresponding to the i-th user in ;

[0086] 104. Determine whether i+1 is greater than K. If so, execute step 105. Otherwise, set i=i+1 and return to step 103.

[0087] 105. Determine whether the number of iterations l reaches the set maximum number of iterations or the algorithm converges, if so, end; otherwise, wait for the central processing unit to return data and repeat steps 102 to 104.

[0088] After the user signal detection is completed through the above process, the estimated value of the transmitted signal is calculated according to the received signal and the channel matrix, and the estimated signal is demodulated to obtain the demodulated signal.

[0089] The method of the present invention is simulated. The bit error rate is used as the performance index of signal detection. The 16QAM modulation and demodulation method is adopted. It is assumed that the channel state is known. Considering that the entire array is divided into 4 sub-arrays, the number of antennas in each sub-array architecture is evenly distributed, that is, N = C × N c (C=4), the total number of base station antennas is N=1024, the total number of users at the same angle but different distances is K=12, and the number of non-'visible' users in each subarray is K / 4, and the center frequency is f c=100GHz, the reference angle of all users arriving at the center of the array is θ=π / 6, the distance from the user to the center of the array is randomly selected, and its range is r∈[1,20]. The number of simulation iterations is 3. PW represents plane wavefront, and SW represents spherical wavefront. The performance of the EP-OSIC algorithm based on the distributed architecture is tested in the system with or without distance parameter modeling, and the performance is compared with the system based on the EP algorithm based on the distributed architecture under the distance parameter modeling. The simulation results are shown in Figure 4 As shown, it can be found that when the system cannot utilize distance information, that is, the system is modeled by a plane wavefront, the signal detection performance is extremely poor, causing the system to lose communication capability. In the spherical wavefront system, that is, considering the influence of the distance parameter on the phase, the EP-SIC algorithm based on a distributed architecture is better than the EP algorithm. This is because when the distance between users is very close, there is interference between users in the system, which causes the performance of the EP algorithm based on a distributed architecture to be worse than that of the EP-OSIC algorithm based on a distributed architecture.

[0090] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting near-field spherical wave ultra-large-scale MIMO uplink signals, characterized in that: To address the problem of large interference between users in a scenario where users are densely populated in the near field, uplink signal detection specifically includes the following steps: The receiving antenna array is divided into C non-overlapping sub-arrays, and each array is used as a factor node and each user's transmitted signal is used as a variable node to construct a factor graph model; According to the factor graph model, the approximate posterior distribution of a signal sent by a certain user on a subarray is calculated; Based on the sorting interference elimination algorithm, each subarray receiving signal and channel matrix are iteratively updated to complete signal detection.

2. The method for detecting a near-field spherical wave ultra-large-scale MIMO uplink signal according to claim 1, characterized in that: In the factor graph model, the process of information transfer between factor nodes and variable nodes is expressed as: in, represents the symbol x sent by the jth user j The corresponding variable node to the cth subarray f c The information transmitted by the corresponding factor node; represents the symbol x sent by the j'th user j' The corresponding variable node to the cth subarray f c The information transmitted by the corresponding factor node, j'∈{1,2,...,K}\j, K is the number of users; Represents the symbol x sent from the variable node f0 corresponding to the prior information to the jth user j The corresponding variable node message; Represents the cth subarray f c The symbol x sent by the corresponding factor node to the jth user j The information transmitted by the corresponding variable node; Represents the c′th subarray f c′ The symbol x sent by the corresponding factor node to the jth user j The information transmitted by the corresponding variable node, c′∈{1,2,...,C}\c, {1,2,...,C}\c represents the set of {1,2,...,C} except c; x\x j represents the symbol set x sent by all users except the symbol x sent by the jth user j The symbol combination that constitutes c (x j ) represents the user x in the probability model j The marginal likelihood function of .

3. The method for detecting a near-field spherical wave ultra-large-scale MIMO uplink signal according to claim 2, characterized in that: Approximate posterior distribution of the jth detected symbol in the cth subarray It is expressed as: Among them, b(x j ) represents the true posterior distribution of the jth detection symbol; represents the exponential family distribution; KL[A‖B] is the KL divergence operation, which is used to characterize the degree of deviation between distribution A and distribution B; represents the symbol x sent by the jth user j The corresponding variable node to the cth subarray f c The information transmitted by the corresponding factor node; Represents the cth subarray f c The symbol x sent by the corresponding factor node to the jth user j The information transmitted by the corresponding variable node.

4. The method for detecting a near-field spherical wave ultra-large-scale MIMO uplink signal according to claim 3, characterized in that: The approximate posterior distribution of all detected symbols in the cth subarray The mean and variance of are expressed as: Among them, μ c represents the approximate posterior distribution of all detected symbols received by the cth subarray The mean of c represents the approximate posterior distribution of all detected symbols in the cth subarray The variance of ; σ represents the noise variance; Denotes the c-th subarray channel matrix H c The transposed matrix of y c represents the received signal of the cth subarray; τ c represents the inverse of the prior variance of all detected symbols received by the cth subarray, γ c represents the a priori mean of all detected symbols received by the c-th subarray.

5. The method for detecting near-field spherical wave ultra-large-scale MIMO uplink signals according to claim 4, characterized in that: Based on the sorting interference elimination algorithm, iteratively updating each subarray receiving signal and channel matrix to eliminate the interference between users includes the following steps:

101. Initialize the number of iterations to l = 0 and calculate the approximate posterior distribution b of all detected symbols in the cth subarray fc The mean and variance of (x); 102. Let l=l+1, calculate the interference-to-signal ratio of each user in each subarray in the lth iteration, and arrange the detection order of users in each subarray according to the ascending sequence of the value to obtain the detection order of the cth subarray And let i=1; 103. Select the test order The i-th user in is taken as the user to be estimated, and the received signal of the c-th subarray and the channel matrix of the c-th subarray are reconstructed using the mean of the approximate posterior distribution of all detected symbols of the i-th user in the c-th subarray; 104. Determine whether i+1 is greater than K. If so, execute step 105. Otherwise, set i=i+1 and return to step 103.

105. Determine whether the number of iterations l reaches the set maximum number of iterations or the algorithm converges, if so, end; otherwise, wait for the central processing unit to return data and repeat steps 102 to 104.

6. The method for detecting near-field spherical wave ultra-large-scale MIMO uplink signals according to claim 5, characterized in that: Reconstructing the received signal of the c-th subarray and the channel matrix of the c-th subarray by using the mean of the approximate posterior distribution of all detected symbols of the i-th user in the c-th subarray includes: in, Indicates that the signal of the i-th user in the detection sequence is calculated The received signal of the c-th subarray obtained by post-reconstruction is equal to the received signal of the c-th subarray when i=1, that is, represents the detection order in the channel matrix of the cth subarray at the i-th iteration The element corresponding to the i-th user in ; Indicates the detection order of the cth subarray in the i-th iteration The mean of the approximate posterior distribution corresponding to the i-th user in .

7. The method for detecting a near-field spherical wave ultra-large-scale MIMO uplink signal according to claim 6, characterized in that: The calculation of the interference-to-signal ratio of each user in each subarray includes: Among them, JSR c,k represents the interference-to-signal ratio of the kth user in the cth subarray; h c,i represents the i-th column of the c-th subarray channel matrix, ||·|| represents the Euclidean norm, and K represents the number of users.

Images