Co-channel interference MIMO detection method based on BP algorithm

Abstract

The invention discloses a co-channel interference MIMO (Multiple Input Multiple Output) detection method based on a BP (Back Propagation) algorithm, relates to the field of digital signal processing, and solves the problems that the bit error rate is relatively high in an interference environment and the calculation complexity exponentially increases although the performance of nonlinear detection is excellent. An interference scene is converted into a standard white noise model, and low-complexity detection is realized through a BP algorithm; or constructing an expansion factor graph model, performing joint probability inference on the interference signal and the target signal, processing continuous distribution variables in combination with an AMP algorithm, and realizing complexity optimization under a low information condition of only needing an interference sample. According to the method, the problem of balancing interference suppression and detection efficiency in a large-scale MIMO system is solved, a high-performance and low-complexity anti-interference detection scheme is provided for a mobile communication system, and the communication quality is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of digital signal processing, and more particularly to a co-channel interference MIMO detection method based on a BP algorithm. Background Art

[0002] MIMO technology is considered a key technology for modern wireless communication systems, particularly fifth-generation (5G) mobile communication systems. As system capacity requirements continue to grow, the application of massive MIMO systems with a high number of antenna elements (AEs) is a key technology path. However, this technology faces the challenge of neighboring cell interference in uplink MIMO detection.

[0003] In existing technologies, maximum likelihood (ML) detectors can achieve optimal performance, but their complexity increases exponentially with the number of transmit antennas, making their implementation increasingly difficult. Other nonlinear detection methods based on QR decomposition, such as sphere decoding and the K-best algorithm, face similar challenges. Linear detection, such as zero-forcing detection and MMSE detection, uses matrix equalization to statistically minimize the objective function. While linear detection has significantly lower complexity, its performance still lags far behind ML detection.

[0004] In practical communication systems, co-channel interference caused by users in adjacent cells must be considered. Existing literature demonstrates that channel capacity is related to the autocorrelation matrix of interference. Therefore, ignoring the correlation between interference and considering only interference power will result in a significant loss of channel capacity. Therefore, it is necessary to study MIMO detection methods in the presence of interference. In 5G systems, the autocorrelation matrix of interference can be estimated using the demodulation reference signal (DMRS). By further considering the autocorrelation matrix of interference, MMSE-IRC can effectively mitigate the impact of interference. As a linear detection algorithm, its performance lags far behind ML and nonlinear detection methods. However, research on BP detection methods for MIMO models with interference is limited, and other nonlinear detection methods also face the problem of high complexity.

[0005] Therefore, how to research and design a MIMO detection method based on the BP algorithm under co-channel interference that can overcome the above-mentioned defects is a problem that we urgently need to solve. Summary of the Invention

[0006] To address the deficiencies in the prior art, the present invention aims to provide a co-channel interference MIMO detection method based on the BP algorithm. Two technical solutions are provided based on practical situations: two structures and algorithms using an expanded factor graph and one using an unexpanded factor graph. If the system has already obtained the autocorrelation matrix of the interference, or only wishes to replace the MIMO detection portion, the unexpanded factor graph structure and algorithm can be used. This solution decomposes the autocorrelation matrix of the interference and transforms the MIMO model under co-channel interference conditions into a standard Gaussian white noise MIMO model, which can then be detected using the standard BP algorithm. If further complexity reduction is desired and the system only obtains interference samples, the expanded factor graph structure and algorithm can be used. This solution simultaneously detects the interference signal using the expanded factor graph.

[0007] The above technical objectives of the present invention are achieved through the following technical solutions:

[0008] A co-channel interference MIMO detection method based on the BP algorithm is provided, comprising the following steps:

[0009] Based on the received signal model and interference noise information, a MIMO model under co-channel interference is constructed;

[0010] defining an interference autocorrelation matrix, and obtaining a channel matrix of virtual noise by decomposing the interference autocorrelation matrix;

[0011] De-correlating the interference of the MIMO model based on the channel matrix of the virtual noise to obtain a decorrelated standard Gaussian white noise MIMO model;

[0012] According to the standard Gaussian white noise MIMO model, a non-expanded factor graph is constructed, wherein the factor graph consists of variable nodes and observation nodes;

[0013] Based on the unexpanded factor graph, the GAI-BP algorithm is used to iteratively transfer soft information to obtain a detection result.

[0014] Furthermore, the expression of the MIMO model under co-channel interference is:

[0015]

[0016] Where y is N r ×1 scale received signal vector, P x is the user's transmission power, H is N r ×N t The channel matrix of scale, x is N t ×1 scale transmitted signal vector, n is N r The power of ×1 scale is σ 2 Gaussian white noise vector, P I is the power of the interfering user, HI Yes N r ×N I The channel matrix of the co-channel interfering user with the same size, x I Yes N I ×1 scale, power normalized, independent interference signal vector, u is the sum of interference and noise.

[0017] Furthermore, the interference autocorrelation matrix is estimated by demodulating the reference signal, and the reversibility of the interference autocorrelation matrix is ensured by adding a correction term to the sample covariance matrix.

[0018] Furthermore, the standard Gaussian white noise MIMO model is obtained by multiplying the MIMO model by the inverse matrix of the channel matrix of the virtual noise:

[0019]

[0020] y'=H'x+n';

[0021] Among them, L -1 is the inverse matrix of the channel matrix of virtual noise, y is the received signal vector, P x is the user's transmit power, H is the channel matrix, x is the transmitted signal vector, y′ is the received signal after decorrelation, H′ is the channel matrix after decorrelation, and n′ is the virtual noise vector of the standard complex Gaussian distribution with mean 0 and variance 1.

[0022] Furthermore, the GAI-BP algorithm is a Gaussian approximation iterative algorithm, which iteratively updates the probability distribution of variable nodes and observation nodes until convergence by assuming that the interference term obeys Gaussian distribution.

[0023] Furthermore, the method further comprises:

[0024] When the system cannot obtain the autocorrelation matrix of the interference, an extended factor graph scheme is adopted, which includes the following steps:

[0025] Based on the received signal model and interference noise information, a MIMO model under co-channel interference is constructed;

[0026] Obtaining an estimated virtual noise channel matrix according to the interference and noise samples, and decorrelating the MIMO model according to the estimated virtual noise channel matrix to obtain a standard Gaussian white noise MIMO model;

[0027] According to the standard Gaussian white noise MIMO model, an extended factor graph is constructed, wherein the factor graph consists of variable nodes and observation nodes;

[0028] Based on the expansion factor graph, the AMP algorithm is used to iteratively transfer soft information to obtain a detection result.

[0029] Furthermore, by decomposing u DMRS The channel matrix of the estimated virtual noise is obtained, specifically:

[0030]

[0031] Where L′ is the estimated channel matrix of virtual noise, is the estimated interference autocorrelation matrix, u DMRS is the interference sample matrix obtained by demodulating the reference signal, P is the number of samples, U p N r ×P-sized unitary matrix, Λ p is a P×P diagonal matrix, and V is a P×P unitary matrix.

[0032] Furthermore, in the extended factor graph scheme, the complexity of the channel matrix of the virtual noise estimated by the interference sample is O(N r P 2 +P 3 ), where P is the number of demodulation reference signal samples and satisfies P<N r .

[0033] Furthermore, the message transmission of the n′ part in the expansion factor graph adopts Gaussian distribution approximation, and its mean and variance are updated by AMP algorithm.

[0034] Furthermore, the detection result is combined with low-density parity check decoding to achieve joint detection and decoding.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. The MIMO detection method based on the BP algorithm under co-channel interference of the present invention fills the gap in MIMO detection methods under co-channel interference conditions provided by the BP algorithm, providing a high-performance, low-complexity MIMO detection algorithm, making it possible to apply the BP algorithm to complex wireless communication systems. Different solutions are provided according to the actual system conditions, with good compatibility.

[0037] 2. This invention converts the received signal into a standard Gaussian white noise model by matrix decomposing the co-channel interference model and performing decorrelation. This technology addresses the technical gap that the traditional BP algorithm does not consider interference correlation, making the BP algorithm directly applicable to co-channel interference scenarios. At the same time, through real and imaginary part separation and the GAI-BP algorithm, performance close to ML detection is achieved under both ideal and non-ideal interference estimation conditions. It is also compatible with the existing system framework and can be adapted by simply replacing the detection module.

[0038] 3. To address the problem of irreversible estimation when the sample is insufficient, the present invention proposes an extended factor graph scheme: the interference signal is extended to the factor graph as a hidden variable, the equivalent channel matrix is constructed by decomposing the interference sample, and the message passing of continuous Gaussian variables is carried out in combination with the AMP algorithm. This scheme breaks through the complexity bottleneck of traditional matrix inversion, realizes low-complexity detection when P << Nr, reduces the time delay, and solves the contradiction between the high complexity of linear detection and the difficulty of practical application of non-linear detection in large-scale MIMO. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, form a part of this application, and do not limit the embodiments of the present invention. In the drawings:

[0040] Figure 1 is the MIMO model under co-channel interference in the embodiments of the present invention.

[0041] Figure 2 is the factor graph structure of the non-extended factor graph scheme in the embodiments of the present invention.

[0042] Figure 3 is the factor graph structure of the extended factor graph scheme in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] To make the purpose, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in combination with embodiments and drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and do not limit the present invention.

[0044] Embodiment: A co-channel interference MIMO detection method based on the BP algorithm.

[0045] Consider a MIMO receiver with N r receive antennas and N t transmit antennas of co-channel users. Assuming that the MIMO channel is an independent and identically distributed flat fading Rayleigh channel, the MIMO model can be modeled as:

[0046]

[0047] where y is a received signal vector with a size of N r ×1, P x is the transmit power of the user, H is a channel matrix with a size of N r ×N t ×1, x is a modulated, power-normalized, and independent transmit signal vector with a size of N t ×1 and M -QAM , and n is a noise vector with a size of N r ×1 and a power of σ2 Gaussian white noise vector. Each element in H obeys an independent standard complex Gaussian distribution with mean 0 and variance 1.

[0048] Further consider the co-channel interference problem in a real system, such as Figure 1 (a) shows that users at the edge of a cell may access other cells, but the base stations of other cells do not know the information of this user, resulting in unknown interference. The simplified model is as follows: Figure 1 (b). Therefore, assuming that N I The above model can be modeled as follows:

[0049]

[0050] Among them, P x is the user's transmit power, H is N r ×N t The channel matrix of size, x is N t ×1 scale transmitted signal vector, P I is the power of the interfering user, H I Yes N r ×N I The channel matrix of the co-channel interfering user with the same size, x I Yes N I ×1 scale, power normalized, independent interference signal vector, u is the sum of interference and noise. Similarly, where H I Each element in follows an independent standard complex Gaussian distribution with mean 0 and variance 1, and n is N r The power of ×1 scale is σ 2 Gaussian white noise vector.

[0051] The autocorrelation matrix of the interference is thus defined as:

[0052]

[0053] Among them, R uu is the interference autocorrelation matrix, The representative size is N r ×N r The identity matrix, u is the sum of interference and noise, P I is the power of the interfering user, x I Yes N I ×1 scale, power normalized, independent interference signal vector, n is N r The power of ×1 scale is σ 2 Gaussian white noise vector. H I is the channel matrix of the interfering transmission signal, σ 2is the power of the Gaussian white noise vector n.

[0054] In systems such as 5G, samples of u can be obtained through demodulation reference signals (DMRS) because these signals are known. Assuming there are P DMRS, then:

[0055] y DMRS =[y1,y2,…,y P ]

[0056] x DMRS =[x1,x2,…,x P ]

[0057] u DMRS =y DMRS -Hx DMRS .

[0058] Therefore the autocorrelation matrix of the interference can be estimated as:

[0059]

[0060] Among them, u DMRS is the interference sample matrix obtained by demodulating the reference signal, is the estimated interference autocorrelation matrix, and P is the number of demodulation reference signals.

[0061] In the system, P<N r , that is, the number of DMRS is less than the number of receiving antennas. In order to make the autocorrelation matrix of interference reversible, it is corrected to:

[0062]

[0063] Among them, α is the regularization parameter, For scale N r ×N r The identity matrix, is the estimated autocorrelation matrix after regularization, is the autocorrelation matrix of the estimated interference.

[0064] According to the central limit theorem, when the number of interfering antennas is large enough, the interference term can be regarded as a Gaussian distribution. Considering that the mean of the interference symbol is 0 and the variance is 1, we have:

[0065]

[0066] Among them, x I Yes N I ×1 scale, power normalized, independent interference signal vector, H Iis the channel matrix of the interfering transmission signal.

[0067] Considering that the noise n is also Gaussian distributed, then:

[0068] u=H I x I +n~CN(0,R uu );

[0069] Where u is the sum of interference and noise, x I Yes N I ×1 scale, power normalized, independent interference signal vector, n is N r The power of ×1 scale is σ 2 Gaussian white noise vector, R uu is the autocorrelation matrix of the interference.

[0070] According to the properties of Gaussian distribution, the correlation between joint Gaussian distributions can be characterized as the linear correlation of independent Gaussian distributions, that is:

[0071] u=Ln';

[0072] Where L is the channel matrix of virtual noise, n' is the virtual noise vector of standard complex Gaussian distribution with mean 0 and variance 1, and:

[0073] E[uu H ]=E[(Ln')(Ln') H ]

[0074] =LL H =R uu ;

[0075] Where u is the sum of interference and noise, L is the channel matrix of virtual noise, n' is the virtual noise vector of standard complex Gaussian distribution with mean 0 and variance 1, R uu is the interference autocorrelation matrix.

[0076] Solution 1: No factor graph expansion

[0077] Assume that the system has obtained R uu The ideal value or estimated value of R uu If the ideal value or estimated value of is reversible, then L must also be reversible. The original system model can be written as:

[0078]

[0079] Where y is N r ×1 scale received signal vector, P x is the user's transmit power, H is N r ×N tThe channel matrix of size, x is N t ×1 scale, M -QAM The modulated and power-normalized, independent transmitted signal vector, L is the channel matrix of virtual noise, and n' is a virtual noise vector with a standard complex Gaussian distribution of mean 0 and variance 1. Each element in H follows an independent, standard complex Gaussian distribution with mean 0 and variance 1.

[0080] Multiply both sides by the inverse matrix of L to remove the correlation of the interference terms, and we have:

[0081]

[0082] y'=H'x+n';

[0083] Among them, y' is the received signal after decorrelation, H' is the channel matrix after decorrelation, L -1 is the inverse matrix of the channel matrix L of the virtual noise.

[0084] It can be found that the MIMO model under co-channel interference is converted to the MIMO model under standard Gaussian white noise, so the GAI-BP algorithm can be used on this basis. To reduce the complexity, the real and imaginary parts can be separated:

[0085]

[0086] in,

[0087]

[0088] in, is the received signal after decorrelation with the real and imaginary parts separated, is the decorrelated channel matrix after separation of real and imaginary parts, is the virtual noise vector separated by real and imaginary parts, is the transmitted signal separated by real and imaginary parts. Obeys a real Gaussian distribution with a mean of 0 and a variance of

[0089] Then the symbol received by the i-th receiving antenna is:

[0090]

[0091] in, is the ith received signal after decorrelation of the real and imaginary parts, is the element in row i and column j of the channel matrix after decorrelation of the real and imaginary parts, is the jth transmitted signal separated by real and imaginary parts, N t is the number of transmitting antennas, is the element in row i and column k of the decorrelated channel matrix after separation of real and imaginary parts, is the ith virtual noise separated from the real and imaginary parts.

[0092] Similarly, according to the central limit theorem, when N t When it is larger, z ij Close to Gaussian distribution, that is:

[0093] z ij ~N(μ ij ,σ ij ),

[0094]

[0095] Among them, z ij is the signal received by the i-th receiving antenna after the real and imaginary parts are separated, except for the signal from the j-th transmitting antenna, μ ij z ij The mean of is the jth transmitted signal separated by real and imaginary parts, N t is the number of transmitting antennas, is the element in row i and column k of the channel matrix after decorrelation of the real and imaginary parts, σ ij for z ij , σ′ is the variance of the virtual noise vector separated by real and imaginary parts.

[0096] In GAI-BP detection, a standard Gaussian white noise MIMO model can be expressed as Figure 1 The factor graph shown is composed of variable nodes (VN) and observation nodes (ON). The soft information M of the lth iteration is l Propagate along the edge of the factor graph, which represents the probability of each constellation point. We define

[0097] is the information transmitted from the i-th VN node to the j-th ON node,

[0098] is the information transmitted from the i-th ON node to the j-th VN node, and s represents M -QAM After the separation of the real and imaginary parts of the constellation points -Ask constellation point. Then:

[0099]

[0100] in, It is the prior probability or the external information transmitted when the channel code is decoded, s k is the kth constellation point, is the ith transmitted signal separated by real and imaginary parts, is the ith received signal after decorrelation of the real and imaginary parts, μ ij z ij The mean of is the element in row i and column k of the channel matrix after decorrelation of the real and imaginary parts, σ ij for z ij The variance, N r is the number of receiving antennas, where z ij is the signal received by the i-th receiving antenna after the real and imaginary parts are separated, except for the signal from the j-th transmitting antenna.

[0101] The BP algorithm produces an approximate posterior probability for each possible constellation point of each transmitted symbol, known as a probability mass function. This result is obtained using iterative posterior probability inference based on Bayesian inference. The variables in the posterior probability function are all discrete. By iteratively updating the approximate posterior probabilities of the VN and ON nodes, they gradually converge to the true posterior probability.

[0102] This solution has excellent performance, requires little modification to the system, and has good compatibility.

[0103] Solution 2: Extended Factor Graph Solution

[0104] In the above method, it is necessary to decompose the matrix R uu And inverse L, the complexity is The complexity is high. To reduce the complexity, we propose a solution of expanding factor graph. This solution further considers that the actual system cannot obtain R uu The ideal value of is often estimated by the sample of u, so we can also directly obtain the L matrix from the sample of u. We assume that there is P(P<N r ) samples, then:

[0105]

[0106] Where y is the received signal vector, P x is the user's transmit power, H is the channel matrix, x is the transmitted signal vector, L is the channel matrix of virtual noise, n′ is the virtual noise vector of the standard complex Gaussian distribution with mean 0 and variance 1, is the estimated autocorrelation matrix after regularization, is the estimated autocorrelation matrix of interference, L′ is the estimated channel matrix of virtual noise, α is the regularization parameter, is the equivalent channel matrix after the expansion factor graph, is the equivalent transmitted signal vector after the expansion factor graph, is the equivalent Gaussian white noise vector after the expansion factor graph. Considering:

[0107]

[0108] Where L′ is the estimated channel matrix of virtual noise, is the estimated interference autocorrelation matrix, u DMRS is the interference sample matrix obtained by demodulating the reference signal, U P is an N r ×P-sized unitary matrix, Λ P Is a P×P diagonal matrix, V is a P×P unitary matrix. Since the decomposition at this time is u DMRS Matrix instead of R uu Matrix, the complexity of the L' matrix is about O(N r P 2 +P 3 ), when P<<N r When the complexity is much smaller than that of the first solution and the traditional solution

[0109] After the conversion, we convert the model to the standard Gaussian white noise MIMO model again, and its factor graph reference Figure 3 shown, but The prior distribution of the n' part is a continuous Gaussian distribution, not a discrete constellation point, so the GAI-BP algorithm cannot be directly applied.

[0110] To solve this problem, we need to introduce a message passing algorithm for continuous variables. The AMP (Approximate Message Passing) algorithm is a standard message passing algorithm for continuous Gaussian variables. It treats all messages as Gaussian distributed. Since the Gaussian distribution can be uniquely determined by its mean and variance, only the mean and variance need to be passed during transmission.

[0111] Similar to the scheme without expanding the factor graph, the symbol received by the i-th receiving antenna is:

[0112]

[0113] in, is the ith received signal separated by real and imaginary parts, is the equivalent channel matrix after the expansion factor graph The element in row i and column j of is the jth equivalent transmitted signal after the expansion factor graph, is the equivalent channel matrix after the expansion factor graph The element in row i and column k of tis the number of transmitting antennas, P is the number of demodulation reference signals, is the signal received by the i-th receiving antenna except the signal of the j-th transmitting antenna after the real and imaginary parts are separated after the expansion factor graph.

[0114]

[0115] in, is the signal received by the i-th receiving antenna except the signal from the j-th transmitting antenna after the real and imaginary parts are separated after the expansion factor graph, for The mean of for The variance of is the kth equivalent transmitted signal after the expansion factor graph.

[0116] for The update of the x part can refer to the non-expanded factor graph scheme, but the mean and variance of the n' part are required, and the mean and variance of this part can be obtained by the formula of the AMP algorithm, namely:

[0117]

[0118] in,

[0119]

[0120] in, is the information transmitted from the i-th ON node to the j-th VN node, is the information transmitted from the i-th VN node to the j-th ON node, N r is the number of receiving antennas, r ij for The mean value, v ij for The variance of for The mean of for , σ′ is the variance of the virtual noise vector separated by real and imaginary parts.

[0121] The AMP algorithm approximates the posterior distribution of continuous variables to a Gaussian distribution. The AMP algorithm's detection results are the mean and variance of this approximate Gaussian distribution for each virtual noise. The detection results are derived using iterative posterior probability inference based on Bayesian inference. The variables in the posterior probability function are all continuous variables. By iteratively updating the approximate posterior probabilities of VN and ON nodes, they gradually converge to the true posterior probabilities.

[0122] Working principle: The present invention is based on the noise characteristics of the MIMO model under co-channel interference. Through the interference autocorrelation matrix decomposition technology (the interference autocorrelation matrix is decomposed and the MIMO model under co-channel interference is decorrelated to obtain a standard Gaussian white noise model) and the dynamic adaptation mechanism of the expanded / unexpanded factor graph structure, combined with the Gaussian approximation iteration of the discrete signal by the GAI-BP algorithm and the mean square error transfer of the continuous interference variable by the AMP algorithm, the interference model is converted into an equivalent white noise model and a hierarchical factor graph is constructed. While retaining the linear complexity advantage of the BP algorithm, the influence of interference correlation is eliminated through the joint probability propagation of the interference signal and the noise, achieving approximation to ML detection performance under ideal estimation (the bit error rate is reduced by 1 order of magnitude compared with MMSE-IRC), and maintaining low complexity operation when the sample is insufficient, finally achieving a high-precision, low-complexity, and highly compatible co-channel interference suppression effect.

[0123] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0124] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0125] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0126] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0127] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A co-channel interference MIMO detection method based on BP algorithm is characterized by: The following steps are involved: Based on the received signal model and interference noise information, a MIMO model under co-channel interference is constructed; defining an interference autocorrelation matrix, and obtaining a channel matrix of virtual noise by decomposing the interference autocorrelation matrix; De-correlating the interference of the MIMO model based on the channel matrix of the virtual noise to obtain a decorrelated standard Gaussian white noise MIMO model; According to the standard Gaussian white noise MIMO model, a non-expanded factor graph is constructed, wherein the factor graph consists of variable nodes and observation nodes; Based on the unexpanded factor graph, the GAI-BP algorithm is used to iteratively transfer soft information to obtain a detection result.

2. The co-channel interference MIMO detection method based on the BP algorithm according to claim 1, characterized in that: The expression of the MIMO model under co-channel interference is: Where y is N r ×1 scale received signal vector, P x is the user's transmission power, H is N r ×N t The channel matrix of scale, x is N t ×1 scale transmitted signal vector, n is N r ×1 The power of scale is σ 2 Gaussian white noise vector, P I is the power of the interfering user, H I N r ×N I The channel matrix of the co-channel interfering user with the same size, x I N I ×1 scale, power normalized, independent interference signal vector, u is the sum of interference and noise.

3. The co-channel interference MIMO detection method based on BP algorithm according to claim 1, characterized in that: The interference autocorrelation matrix is estimated by demodulating a reference signal, and the reversibility of the interference autocorrelation matrix is ensured by adding a correction term to a sample covariance matrix.

4. The co-channel interference MIMO detection method based on BP algorithm according to claim 1, characterized in that: The standard Gaussian white noise MIMO model is obtained by multiplying the MIMO model by the inverse matrix of the channel matrix of the virtual noise: y'=H'x+n'; Among them, L -1 is the inverse matrix of the channel matrix of virtual noise, y is the received signal vector, P x is the user's transmit power, H is the channel matrix, x is the transmitted signal vector, y′ is the received signal after decorrelation, H′ is the channel matrix after decorrelation, and n′ is the virtual noise vector of the standard complex Gaussian distribution with mean 0 and variance 1.

5. The co-channel interference MIMO detection method based on BP algorithm according to claim 1, characterized in that: The GAI-BP algorithm is a Gaussian approximation iterative algorithm that iteratively updates the probability distribution of variable nodes and observation nodes until convergence by assuming that the interference term obeys a Gaussian distribution.

6. The co-channel interference MIMO detection method based on BP algorithm according to claim 1, characterized in that: The method further includes: When the system cannot obtain the autocorrelation matrix of the interference, an extended factor graph scheme is adopted, which includes the following steps: Based on the received signal model and interference noise information, a MIMO model under co-channel interference is constructed; Obtaining an estimated virtual noise channel matrix according to the interference and noise samples, and decorrelating the MIMO model according to the estimated virtual noise channel matrix to obtain a standard Gaussian white noise MIMO model; According to the standard Gaussian white noise MIMO model, an extended factor graph is constructed, wherein the factor graph consists of variable nodes and observation nodes; Based on the expansion factor graph, the AMP algorithm is used to iteratively transfer soft information to obtain a detection result.

7. The co-channel interference MIMO detection method based on BP algorithm according to claim 6, characterized in that: By decomposing u DMRS The channel matrix of the estimated virtual noise is obtained, specifically: Where L′ is the estimated channel matrix of virtual noise, is the estimated interference autocorrelation matrix, u DMRS is the interference sample matrix obtained by demodulating the reference signal, P is the number of samples, U p N r ×P-sized unitary matrix, Λ p is a P×P diagonal matrix, and V is a P×P unitary matrix.

8. The co-channel interference MIMO detection method based on BP algorithm according to claim 6, characterized in that: In the expansion factor graph scheme, the complexity of the channel matrix of the virtual noise estimated by the interference sample is O(N r P 2 +P 3 ), where P is the number of demodulation reference signal samples and satisfies P<N r .

9. The co-channel interference MIMO detection method based on BP algorithm according to claim 6, characterized in that: The message transmission of the n′ part in the expansion factor graph adopts Gaussian distribution approximation, and its mean and variance are updated by AMP algorithm.

10. The co-channel interference MIMO detection method based on BP algorithm according to claim 1, characterized in that: The detection result is combined with low-density parity check decoding to achieve joint detection and decoding.

Images