Information geometry approach to massive MIMO signal detection
Through information geometry method, the target manifold and auxiliary manifold are constructed in a large-scale MIMO system, and the m-projection is iteratively calculated to update the probability distribution, solving the problems of high complexity and performance degradation of existing signal detection algorithms, and achieving high performance Signal detection and acquisition of edge posterior probability distribution with low complexity.
Patent Information
- Application Number
- CN202310741285.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-21
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2043-06-21
AI Technical Summary
In large-scale MIMO systems, existing signal detection algorithms such as maximum posterior detection and linear minimum mean square error detectors have problems with high computational complexity and degradation of performance, especially under high-order modulation conditions, it is difficult to achieve high complexity in low complexity. Performance signal detection.
Using information geometry method, by constructing target manifolds and auxiliary manifolds, iteratively calculates m-projection to update the probability distribution, and obtains the edge posterior probability distribution of symbols sent by each user, thereby demodulating, reducing the computational complexity and improving code errors rate performance.
The system bit error rate performance is significantly improved, and the calculation complexity is further reduced compared with the existing technical means, achieving high-performance signal detection.
Smart Images

Figure CN116599560B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of communication technology and relates to an information geometry method for large-scale MIMO signal detection. Background Art
[0002] As a key enabling technology for the fifth generation (5G) mobile communication system, massive multiple input multiple output (MIMO) technology can bring significant gains to the spectrum efficiency and energy efficiency of the communication system. With the further increase in the number of antennas, massive MIMO will continue to play a key role in the future sixth generation (6G) mobile communication system. In order to give full play to the various advantages of massive MIMO systems, signal detection is crucial. Given a received signal, the task of the detector at the receiving end is to determine the transmitted symbols of each user. The optimal detector based on the maximum a posteriori criterion or maximum likelihood criterion performs an exhaustive search and obtains the transmitted symbols of each user. Optimal detection is a non-deterministic polynomial time problem, and its computational complexity grows exponentially with the number of values of the symbols to be detected, so its application in massive MIMO systems is limited. On the other hand, linear detectors such as the linear minimum mean squared error (LMMSE) detector are widely used due to their polynomial time level complexity. However, theoretical studies have shown that the estimation of the transmitted symbols by the LMMSE detector is biased, and at the same time, the performance of the LMMSE detector will be seriously degraded in massive MIMO systems with high-order modulation. Therefore, the design of low-complexity and high-performance detectors is an urgent problem to be solved in massive MIMO signal detection.
[0003] The signal detection problem can be regarded as a statistical inference problem under discrete random variables. The space defined by the parameters of the posterior probability distribution of the transmitted symbols can be regarded as a differentiable manifold with a Riemann structure, and its Riemann measure is given by the Fisher information matrix. Therefore, the definitions and tools in differential geometry can be well applied to statistical inference problems, and it is appropriate and reasonable to apply information geometry to signal detection. The main idea of information geometry is to study the intrinsic geometric structure of a specific set of probability distributions by treating the parameter space of probability distributions as differentiable manifolds. In recent years, information geometry has been successfully applied to massive MIMO channel estimation, false alarm rate detection, and arrival angle estimation. Using information geometry theory to study signal detection problems has the following advantages: first, information geometry can provide a unified framework for the theoretical analysis of existing algorithms; second, information geometry theory provides an intuitive understanding of statistical models from a geometric perspective, which can promote the intrinsic study of existing problems; in addition, information geometry can also improve existing algorithms from a more general and intrinsic perspective. Summary of the invention
[0004] Technical problem: In response to the shortcomings of the existing technology, the present invention discloses an information geometry method for large-scale MIMO signal detection, which can obtain the marginal a posteriori probability distribution of the symbols sent by each user. While significantly improving the system bit error rate performance, it can further reduce the computational complexity compared with existing similar technical means.
[0005] Technical solution: In order to achieve the above-mentioned purpose, the present invention provides a technical solution of an information geometry method for large-scale MIMO signal detection as follows:
[0006] The base station side obtains the marginal distribution of the posterior probability distribution of the symbols sent by each user by using the information geometry method through the received uplink signal and the channel state information of each user, thereby obtaining the posterior probability information of the symbols sent by each user; the information geometry method defines the set of probability distributions of discrete random vectors as the original manifold, and constructs the target manifold and the auxiliary manifold according to the posterior probability distribution of the sent symbol vector, iteratively calculates the m-projection of the probability distribution in the auxiliary manifold on the target manifold, and updates the probability distribution in the target manifold and the auxiliary manifold according to the m-projection. After the iteration converges, the probability distribution on the target manifold is used as the marginal posterior probability distribution of the symbols sent by each user, and demodulation is performed according to this distribution.
[0007] The target manifold is a set of probability distributions of discrete random vectors whose components are independent of each other.
[0008] The auxiliary manifold is a set of probability distributions of discrete random vectors containing a single cross term.
[0009] The m-projection is obtained by minimizing the KL divergence (Kullback-Leibler divergence) between the probability distribution in the auxiliary manifold and the target manifold. The KL divergence itself is a function. Given two probability distributions, the expression of the KL divergence is determined, and the parameters of a probability distribution can be directly obtained by minimizing this KL divergence.
[0010] The target manifold and the auxiliary manifold, wherein the probability distribution is represented by their respective natural parameters; wherein the probability distribution in the target manifold is represented by first summing up the exponentials of its natural parameters and then taking the inverse thereof; after determining the probability distribution in an auxiliary manifold and the values of its natural parameters, this probability distribution is obtained by traversing the values in the modulation symbol set and then normalizing the probability.
[0011] The step of using the information geometry method to obtain the marginal posterior probability distribution of the symbols sent by each user includes:
[0012] Step 1, establishing the original manifold, target manifold and auxiliary manifold for massive MIMO signal detection;
[0013] Step 2, initialize the natural parameters of the probability distribution in the auxiliary manifold and the target manifold;
[0014] Step 3, calculating the m-projection of the probability distribution in the auxiliary manifold on the target manifold according to the natural parameters of the probability distribution in the auxiliary manifold, the received data signal and the channel state information of each user;
[0015] Step 4, update the natural parameters of the probability distribution in the auxiliary manifold and the target manifold according to the m-projection;
[0016] Step 5, repeat steps 3-4 until the preset number of iterations or the natural parameters of the probability distribution in the target manifold converge, obtain the marginal posterior probability distribution of the symbols sent by each user and complete demodulation.
[0017] After obtaining the marginal a posteriori probability distribution of the symbols sent by each user, the modulation symbol with the largest marginal a posteriori probability is selected from the modulation set as the demodulation result, or soft demodulation is performed using the marginal a posteriori probability distribution to obtain the likelihood ratio information of each code bit.
[0018] Beneficial effects: Compared with the prior art, the information geometry method for massive MIMO signal detection proposed in the present invention can obtain the marginal a posteriori probability distribution of the symbols sent by each user. According to this marginal a posteriori probability distribution, the modulation symbol that maximizes the marginal a posteriori probability distribution can be selected from the modulation set as the demodulation result, or the marginal a posteriori probability distribution can be used for soft demodulation to further obtain the likelihood ratio information of each code bit. While significantly improving the system bit error rate performance, the computational complexity can be further reduced compared to existing similar technical means such as LMMSE detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 is a flow chart of a signal detection method according to an embodiment of the present invention;
[0020] Figure 2 It is a flow chart of the information geometry method for signal detection in an embodiment of the present invention;
[0021] Figure 3 It is a schematic diagram of comparing the bit error rate performance of the information geometry method under 64-QAM modulation in an embodiment of the present invention and the existing method;
[0022] Figure 4 It is a schematic diagram comparing the convergence curves of the information geometry method and the existing method when the signal-to-noise ratio is 19dB and 64-QAM modulation in an embodiment of the present invention. DETAILED DESCRIPTION
[0023] The technical solution provided by the present invention will be described in detail below in conjunction with specific embodiments. It should be understood that the following specific implementation methods are only used to illustrate the present invention and are not used to limit the scope of the present invention.
[0024] like Figure 1 As shown, an information geometry method for large-scale MIMO signal detection disclosed in an embodiment of the present invention includes obtaining channel state information of each user terminal through channel estimation on the base station side; obtaining the posterior probability information of the symbols sent by each user and completing demodulation by using the information geometry method through the received uplink data signal; wherein the information geometry method defines the set of probability distributions of discrete random vectors as the original manifold, and constructs a target manifold and an auxiliary manifold according to the posterior distribution of the symbols sent by each user, iteratively calculates the m-projection of the probability distribution in the auxiliary manifold on the target manifold, and updates the probability distribution in the target manifold and the auxiliary manifold according to the m-projection, and finally uses the probability distribution in the target manifold as the marginal posterior probability distribution of the symbols sent by each user.
[0025] Figure 2 The specific steps of obtaining the marginal posterior probability distribution using the information geometry method are illustrated, including: (1) establishing the original manifold, auxiliary manifold and target manifold for large-scale MIMO signal detection; (2) initializing the natural parameters distributed in the auxiliary manifold and the target manifold; (3) calculating the m-projection of the probability distribution in the auxiliary manifold on the target manifold based on the natural parameters distributed in the auxiliary manifold, the received data signal and the channel state information of each user; (4) updating the natural parameters distributed in the auxiliary manifold and the target manifold based on the m-projection; repeating steps (3)-(4) until a preset number of iterations or the natural parameters distributed in the target manifold converge, and using the probability distribution in the target manifold as the marginal posterior probability distribution of the symbols sent by each user.
[0026] The method of the present invention is mainly applicable to a large-scale MIMO system in which a large-scale antenna array is equipped at the base station side to simultaneously serve multiple users. The specific implementation process of the information geometry method of signal detection involved in the present invention is described in detail below in conjunction with a specific communication system example. It should be noted that the method of the present invention is not only applicable to the specific system model exemplified below, but also to system models with other configurations.
[0027] 1. System Configuration and Signal Detection Problem Statement
[0028] 1. System Model
[0029] Consider the following massive MIMO system. The base station is equipped with a uniform planar array (UPA) with N antennas. r =N r,v ×N r,h , N r,v and N r,h are the number of columns and rows of the uniform array, respectively. The base station serves K users with single antennas in the same cell at the same time. The received signal model is described in detail below. Send a symbol vector to each user, where is the symbol sent by the kth user, is the modulation symbol set, is the modulation order, is the set of labels for each user. Order QAM modulation. Assume that the average energy of the transmitted symbol is normalized, that is, in represents the expected operation symbol. Assume that the energy of the transmitted symbol of each user is normalized, that is, Send symbol vector After passing through the flat fading complex channel matrix, the received signal vector at the base station side is can be expressed as
[0030]
[0031] in is the channel matrix, is a cyclically symmetric Gaussian noise, and the noise power is In this article and Respectively represent the N×M dimensional complex vector space and the N×M dimensional real vector space. Further, the above complex received signal model is transformed into a real received signal model, and the real vector is defined as as well as in Respectively represent taking the real part and taking the imaginary part, and then define a real matrix:
[0032]
[0033] Then the actual received signal model can be expressed as:
[0034] y=Gs+z, (3)
[0035] in is the real transmitted symbol vector, is the set of values of the real and imaginary parts of the transmitted symbol, is a real Gaussian noise vector, The square root operator symbol.
[0036] 2. Problem Statement
[0037] Given the received signal model (3), the posterior probability distribution of the transmitted vector s can be expressed as:
[0038]
[0039] in Is the sending symbol s k The prior probability, y n is the nth element of y, f(y n |s) is given s, y n The probability density function of It is 2N r The nth column of the dimensional identity matrix. Information geometry methods can approximate the marginal distribution of the posterior probability distribution p(s|y), that is, According to the posterior marginal distribution, the maximum a posteriori detection is performed, that is,
[0040]
[0041] where s k,MPM Yes k The detection value of . Those skilled in the art will understand that the channel matrix in the received signal model (1) In other received signal models, different forms may be used. Any statistical inference problem that satisfies the received signal model (1) and the discrete distribution prior can be solved using information geometry methods, and is not limited to the problems described in this example.
[0042] 2. Information geometry approach to massive MIMO signal detection
[0043] 1. Establishment of original manifold, auxiliary manifold and target manifold for massive MIMO signal detection
[0044] Define the original manifold as the set of probability distributions of the sending vector s, that is,
[0045]
[0046] Next, the posterior probability distribution of the transmission vector s is further expressed as:
[0047]
[0048] in
[0049]
[0050] ln means taking the natural logarithm, and the superscript T means the transposition operator.
[0051]
[0052] ψ q is the normalization factor. Based on the posterior probability distribution, the target manifold is defined as the set of probability distributions of discrete random vectors of each component, that is,
[0053]
[0054]
[0055] in is the natural parameter of p0(s;θ0), θ 0,k =[θ 0,k,1 ,θ 0,k,2 ,…,θ 0,k,L-1 ] T , the superscript T indicates the transposition, ψ0(θ0) and ψ0(θ 0,k ) is the normalization factor (free energy),
[0056]
[0057] Given p0(s;θ0) and its marginal probability distribution p0(s k θ 0,k ), the probability value of the marginal distribution can be explicitly expressed as:
[0058]
[0059] In formula (18), Next, define 2N r auxiliary manifolds, where the nth auxiliary manifold is defined as:
[0060]
[0061] in Yesn (s;θ n )’s natural parameter, θ n,k =[θ n,k,1 ,θ n,k,2 ,…,θ n,k,L-1 ] T , ψ n (θ n ) is the normalization factor (free energy).
[0062] 2. Signal detection method
[0063] The information geometry method of signal detection is divided into four steps, namely, establishing the original manifold, auxiliary manifold and target manifold for massive MIMO signal detection; initializing the natural parameters of the probability distribution in the auxiliary manifold and the target manifold; calculating the m-projection of the probability distribution in the auxiliary manifold on the target manifold based on the natural parameters of the probability distribution in the auxiliary manifold, the received data signal and the channel state information of each user; and updating the natural parameters distributed in the auxiliary manifold and the target manifold based on the m-projection. The step of establishing the original manifold, auxiliary manifold and target manifold for massive MIMO signal detection is described in the previous section, and the subsequent steps are described below.
[0064] (1) Initialize the natural parameters of the probability distribution in the auxiliary manifold and the target manifold
[0065] Set the initial value of the number of iterations t = 0 and initialize the target manifold parameters And auxiliary manifold parameters The superscript t indicates the tth iteration.
[0066] (2) Calculate the m-projection of the probability distribution in the auxiliary manifold on the target manifold
[0067] The m-projection of the probability distribution in the auxiliary manifold on the target manifold is calculated based on the natural parameters of the probability distribution in the auxiliary manifold, the received data signal, and the channel state information of each user. Its m-projection on the target manifold is The m-projection is to find a distribution p0(s; θ 0n ), so that p0(s;θ 0n ) and the distribution p in the auxiliary manifold n (s;θ n ) has the smallest KL divergence, that is,
[0068]
[0069] in is a constant that is independent of θ0. Theoretical proof: given Its m-projection satisfies the following conditions:
[0070]
[0071] in is a 2K-1 discrete random vector obtained by removing the kth element in s, p n (s k θ n ) is p n (s;θ n ) is the marginal probability distribution of . Next, the m-projection p0(s; θ 0n ) is expressed as follows:
[0072]
[0073] in According to the central limit theorem, when the number of users is large, ξ n,k It can be approximately expressed as:
[0074] ξ n,k =[ξ n,k,1 ,ξ n,k,2 ,…,ξ n,k,L-1 ] T , (28)
[0075]
[0076] where g n,k is the (n,k)th element of the channel matrix G,
[0077]
[0078] where y n is the nth element of the receiving vector y.
[0079] (3) Update the natural parameters of the probability distribution in the auxiliary manifold and the target manifold
[0080] According to the above process, we can calculate After that, the natural parameters of the probability distribution in the auxiliary manifold and the target manifold are updated as follows:
[0081]
[0082] Where 0<α≤1 is the damping coefficient.
[0083] Further update the number of iterations t = t + 1 and repeat the calculation And update the natural parameters of the probability distribution in the auxiliary manifold and the target manifold until the preset number of iterations or θ0 converges. Given θ0, the marginal distribution of the probability distribution in the target manifold is given by formulas (17) and (18), and then p0(sk θ 0,k ) as the a posteriori marginal probability p(s) of the transmitted symbol k |y) and complete demodulation through formula (5), or perform soft demodulation through marginal posterior probability distribution, where the marginal posterior probability of each user sending symbol is given by the following formula, and the likelihood ratio information of each user code bit can be further determined in combination with the decoding method.
[0084]
[0085]
[0086] The computational complexity of a single iteration of the information geometry method for signal detection is in Represents symbols for quantities of the same order.
[0087] 3. Implementation Effect
[0088] In order to enable those skilled in the art to better understand the solution of the present invention, a comparison of signal detection performance results between the information geometry method in this embodiment and the existing method under a specific system configuration is given below.
[0089] First, the detection performance comparison between the information geometry method in this embodiment and the existing methods is given. The compared methods include LMMSE detection, the EP detection method proposed in the document "Expectation Propagation Detection for High-Order High-Dimensional MIMO Systems, in IEEE Transactions on Communications, vol. 62, no. 8, pp. 2840-2849, Aug. 2014, doi: 10.1109 / TCOMM.2014.2332349." and the AMP algorithm in the document "Message passing algorithms for compressed sensing: I. motivation and construction, 2010 IEEE Information Theory Workshop on Information Theory (ITW2010, Cairo), Cairo, Egypt, 2010, pp. 1-5, doi: 10.1109 / ITWKSPS.2010.5503193.". Consider a configuration of N r =128, K=36 and N t = 1, where the base station antenna configuration is N r,v=8,N r,h =16, the transmitted symbol adopts 64-QAM modulation, and the signal-to-noise ratio (SNR) is defined as The computational complexity of the LMMSE detector is The computational complexity of a single iteration of the EP detector is The computational complexity of a single iteration of AMP detection is Among all algorithms, the computational complexity of EP detector is the highest. When the number of iterations is low (such as dozens of times), the computational complexity of IGA detector is lower than that of LMMSE detector. The computational complexity of AMP detection is the lowest.
[0090] Figure 4 The bit error rate performance comparison of the information geometry method in this embodiment and the EP and AMP methods under different signal-to-noise ratios is given in the considered massive MIMO system. The maximum number of iterations of the information geometry method, EP, and AMP is set to 40. Figure 4 It can be seen that the information geometry method (IGA) can achieve the best bit error rate performance under all signal-to-noise ratios. -2 When , the signal-to-noise ratio gain of IGA compared to EP and AMP is 2dB and 3dB respectively. This shows that compared with the existing method, the information geometry method in this embodiment can obtain better bit error rate performance.
[0091] Next, a schematic diagram comparing the convergence curves of the information geometry method and the existing methods in this embodiment is given. Keeping the large-scale parameters under consideration unchanged, the signal-to-noise ratio is set to 28dB, and using LMMSE detection as the bit error rate baseline for performance reference, the convergence curves of the IGA, EP and AMP methods are drawn. It can be found that the information geometry method (IGA) requires about 50 iterations to converge, and its bit error rate performance is the best after convergence. The EP and AMP methods require 30 and 80 iterations to converge, respectively. This shows that compared with the existing algorithms, the information geometry method can obtain the best bit error rate performance with the lowest computational complexity.
[0092] Based on the same inventive concept, an embodiment of the present invention discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the computer device implements the above-mentioned method for detecting large-scale MIMO signals.
[0093] In a specific implementation, the device includes a processor, a communication bus, a memory, and a communication interface. The processor may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present invention. The communication bus may include a path to transmit information between the above components. The communication interface uses any transceiver-like device for communicating with other devices or communication networks. The memory may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a read-only compact disk (CD-ROM) or other optical disk storage, a disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to this. The memory may be independent and connected to the processor through a bus. The memory may also be integrated with the processor.
[0094] The memory is used to store the application code for executing the scheme of the present invention, and the execution is controlled by the processor. The processor is used to execute the application code stored in the memory, thereby implementing the channel estimation method provided by the above embodiment. The processor may include one or more CPUs, or may include multiple processors, each of which may be a single-core processor or a multi-core processor. The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).
[0095] Based on the same inventive concept, an embodiment of the present invention discloses a massive MIMO communication system, including a base station and multiple user terminals, characterized in that the base station side obtains the marginal distribution (marginal a posteriori probability distribution) of the posterior probability distribution of the symbols sent by each user through the received uplink signal and the channel state information of each user using an information geometry method, thereby obtaining the posterior probability information of the symbols sent by each user. The specific signal detection information geometry method is referred to the aforementioned embodiment and will not be repeated here.
[0096] Based on the same inventive concept, an embodiment of the present invention discloses a large-scale MIMO communication system, including a base station and multiple user terminals, wherein the base station or the user terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is loaded into the processor to implement the aforementioned large-scale MIMO signal detection method. In the embodiments provided in the present application, it should be understood that the disclosed method can be implemented in other ways without exceeding the spirit and scope of the present application. The current embodiment is only an illustrative example and should not be used as a limitation, and the specific content given should not limit the purpose of the present application. For example, some features can be ignored or not executed.
[0097] The technical means disclosed in the scheme of the present invention are not limited to the technical means disclosed in the above-mentioned implementation mode, but also include technical schemes composed of any combination of the above-mentioned technical features. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also regarded as the protection scope of the present invention.
Claims
1. An information geometry method for massive MIMO signal detection, characterized in that: The method comprises the following steps: a base station side obtains the marginal distribution of the posterior probability distribution of symbols sent by each user by using an information geometry method through the received uplink signal and the channel state information of each user, thereby obtaining the marginal distribution information of the posterior probability distribution of symbols sent by each user; the information geometry method defines the set of probability distributions of discrete random vectors as the original manifold, and constructs a target manifold and an auxiliary manifold according to the posterior probability distribution of the sent symbol vector, iteratively calculates the m-projection of the probability distribution in the auxiliary manifold on the target manifold, and updates the probability distribution in the target manifold and the auxiliary manifold according to the m-projection; after the iteration converges, the probability distribution on the target manifold is used as the marginal distribution of the posterior probability distribution of symbols sent by each user, and demodulation is performed according to the marginal distribution of this posterior probability distribution; The above method specifically includes the steps of: Step 1, establishing the original manifold, target manifold and auxiliary manifold for massive MIMO signal detection; Step 2, initialize the natural parameters of the probability distribution in the auxiliary manifold and the target manifold; Step 3, calculating the m-projection of the probability distribution in the auxiliary manifold on the target manifold according to the natural parameters of the probability distribution in the auxiliary manifold, the received data signal and the channel state information of each user; Step 4, update the natural parameters of the probability distribution in the auxiliary manifold and the target manifold according to the m-projection; Step 5, repeating steps 3 and 4 until the preset number of iterations or the natural parameters of the probability distribution in the target manifold converge, obtaining the marginal distribution of the posterior probability distribution of the symbols sent by each user and completing demodulation; The target manifold is a set of probability distributions of discrete random vectors whose components are independent of each other; The auxiliary manifold is a set of probability distributions of discrete random vectors containing a single cross term; The m-projection is obtained by minimizing the KL divergence between the probability distribution in the auxiliary manifold and the target manifold.
2. The information geometry method for massive MIMO signal detection according to claim 1, characterized in that: The target manifold and the auxiliary manifold, wherein the probability distribution is represented by their respective natural parameters; wherein the probability distribution in the target manifold is represented by first summing up the exponentials of its natural parameters and then taking the inverse thereof; after determining the probability distribution in an auxiliary manifold and the values of its natural parameters, the probability distribution in this auxiliary manifold is obtained by traversing the values in the modulation symbol set and then normalizing the probability.
3. The information geometry method for massive MIMO signal detection according to claim 1, characterized in that: After obtaining the marginal distribution of the posterior probability distribution of the symbols sent by each user, the modulation symbol that maximizes the marginal distribution of the posterior probability distribution is selected in the modulation set as the demodulation result, or soft demodulation is performed using the marginal distribution of the posterior probability distribution to obtain the likelihood ratio information of each code bit.
