A blind channel estimation and signal detection method and system for a massive MIMO system

Abstract

The application belongs to the field of information and communication technology, and discloses a blind channel estimation and signal detection method and system of a large-scale MIMO system, wherein channel coefficient matrix and signal matrix are probabilistically modeled at the base station receiving end to obtain the prior distribution of S and X; the number of grid points of a guide vector matrix and the number of bits of a quantizer are designed; the prior distribution is initialized to obtain the initial value of the channel coefficient matrix and the covariance matrix between the columns thereof, and the initial value of the signal matrix and the covariance matrix between the rows thereof; the channel coefficient matrix and the signal matrix are estimated based on the received signal; the permutation ambiguity and the scale ambiguity of the estimated value of the channel coefficient matrix and the estimated value of the signal matrix are determined, and the estimated value of the channel matrix is obtained by calculation. The application develops a generalized matrix decomposition algorithm based on a mixed vector message passing to perform channel estimation and signal detection, has superior performance in channel estimation and signal detection, and achieves lower time complexity.

Description

Technical Field

[0001] This invention belongs to the field of information and communication technology, and particularly relates to a blind channel estimation and signal detection method and system for a large-scale MIMO system. Background Technology

[0002] In recent years, massively multi-input multiple-output (MIMO) systems have been extensively studied due to their advantages in increasing system throughput, link reliability, and energy efficiency. Because multiple antennas transmit data simultaneously, such systems can process multiple data streams in parallel, thus multiplying system throughput. Furthermore, massively multi-input multiple-output (MIMO) systems can further improve system capacity by employing spatial multiplexing techniques and utilize beamforming techniques to enhance signal strength and coverage, thereby ensuring link reliability.

[0003] However, the performance of massive MIMO systems is limited by channel estimation and signal detection. Channel estimation is the process of estimating the characteristics of a wireless channel, while signal detection is the process of detecting the original transmitted signal from the received signal based on the channel estimation. In massive MIMO systems, the large number of antennas makes channel estimation and signal detection even more complex and difficult. Traditional channel estimation methods typically use training sequences, which requires significant time and resources. Blind channel estimation methods, on the other hand, do not require training sequences and can perform channel estimation and signal detection simultaneously by analyzing the statistical characteristics of the received signal. Due to the large number of antennas in massive MIMO systems, using blind channel estimation and signal detection can effectively reduce pilot overhead.

[0004] Due to the limited number of scatterers in the propagation environment, experimental studies have demonstrated the sparse structure of the physical channel angular domain in massive MIMO systems. In the paper J. Zhang, X. Yuan, and Y.-JA Zhang, “Blind signal detection in massive MIMO: Exploiting the channel sparsity,” IEEE Trans. Commun., vol. 66, no. 2, pp. 700-712, 2017, the authors utilized channel sparsity to simultaneously estimate the channel and detection signals from the received signal. By improving the bilinear generalized approximate message passing algorithm, the throughput of massive MIMO systems achieved a significant improvement over previous methods in the high SNR range. In the paper H. Liu, X. Yuan, and YJ Zhang, “Super-resolution blind channel-and-signal estimation for massive MIMO with one-dimensional antenna array,” IEEE Trans. Signal Process., v0l.67, no.17, pp.4433-4448, 2019, the authors further considered the angle mismatch and performance loss problems in DFT-based methods and proposed a novel blind channel estimation and signal detection algorithm based on message passing. However, existing methods all suffer from poor channel and signal estimation performance and high time complexity.

[0005] The closest existing technology and its problems:

[0006] 1) Blind Channel Estimation: This is a promising solution to the "pilot pollution" problem in massive MIMO. Pilot pollution refers to the interference of pilot signal sequences transmitted by users in other cells with the channel estimation in one cell, rendering increases in the number of antennas or transmission power ineffective. Furthermore, a study compared the performance of pilot-based, semi-blind, blind, and adaptive blind channel estimation methods using a cluster-based COST 2100 channel model through MATLAB simulations.

[0007] 2) Deep learning-assisted blind channel estimation: This method utilizes the asymptotic orthogonality of large-scale MIMO channels to average out channel distortion without requiring prior knowledge of the channel impulse response. Through some mathematical operations, data symbols transmitted by different users can be extracted.

[0008] 3) Blind signal detection using channel sparsity: In practical large-scale MIMO systems, most of the system resources are consumed in order to obtain channel state information (CSI), resulting in a significant reduction in system capacity compared to the ideal situation (when perfect CSI is available).

[0009] 4) Semi-blind channel and signal estimation: Considering the channel sparsity in the angular domain, this approach involves the transceiver design of uplink massive MIMO systems.

[0010] Based on these existing technologies, the main technical problems they face include:

[0011] - Pilot pollution: In large-scale MIMO systems, channel estimation is susceptible to interference from other user signals, especially when using pilot-based channel estimation methods.

[0012] - Resource consumption and system capacity limitations: To acquire channel state information (CSI), large-scale MIMO systems require a large amount of system resources, which leads to a significant reduction in system capacity.

[0013] - Utilizing channel sparsity: Effectively utilizing channel sparsity in channel estimation and signal detection is a major challenge, requiring precise algorithms and mathematical processing. Summary of the Invention

[0014] To address the problems existing in the prior art, this invention provides a blind channel estimation and signal detection method and system for large-scale MIMO systems.

[0015] This invention is implemented as follows: a blind channel estimation and signal detection method for a large-scale MIMO system, comprising the following steps:

[0016] Step 1: At the base station receiver, perform probabilistic modeling of the channel coefficient matrix and signal matrix to obtain the prior distributions p(S) and p(X) of S and X; design the steering vector matrix. The number of grid points and the number of bits in the quantizer φ(·);

[0017] Step two: Initialize the channel coefficient matrix according to the prior distribution in step one to obtain the initial values. and the covariance matrix V between its columns s Initial values ​​of the signal matrix and the covariance matrix U between its rows x ;

[0018] Step 3: The GBF-HVMP algorithm is used to estimate the channel coefficient matrix S and the signal matrix based on the received signal y;

[0019] Step 4: Estimate the channel coefficient matrix based on the pilot sequence elimination result. With signal matrix estimate The permutation fuzziness and scale fuzziness are determined, and the results are obtained through calculation. The estimated value of the channel in step one is obtained.

[0020] Furthermore, suppose the base station is configured with a uniform linear array of N receiving antennas to serve K single-antenna users, let θ k,i This represents the i-th angle of arrival of the incoming signal from the k-th user. It is the set of arrival angles.

[0021] Furthermore, the signal angle of arrival is divided into L resolution intervals, where the l-th interval is composed of... In this expression, d is the spacing between any two adjacent receiving antennas, λ is the propagation wavelength, and l∈{0,1,…,L-1}; let Covering angles from -90° to 90°, when When the channel between the k-th user and the base station in the angle domain is represented as:

[0022]

[0023] In the formula, These are sparse channel coefficients in the angular domain. It is a guide vector. The channel is block fading with a coherence time of T, and the channel remains unchanged within each transport block.

[0024] Furthermore, for each coherence time T, the signal of the k-th user is represented as... The set of all transmitted signals from the user terminal is The first N p The sequence is listed as a pilot sequence; the received signal is given by the following formula:

[0025] y = φ(z),

[0026] z = Fw + n,

[0027]

[0028] ω = vec(SX),

[0029] In the formula, φ(·) is the element-wise quantizer generated at the receiver in the MIMO system due to precision limitations. This represents the Kronecker product, vec(·) represents the vectorization operator, and z is the unquantized received signal. The channel coefficient matrix, Let n represent an AWGN vector; the elements of n are independent and identically distributed, and all follow a certain distribution. Distribution, where σ 2 This represents noise power.

[0030] Furthermore, step three first calculates the message. The mean is:

[0031]

[0032] Compute message The mean and variance are respectively:

[0033]

[0034]

[0035] In the formula, The normalization factor; the message m obtained z (z) represents the estimate of the unquantized received signal.

[0036] Furthermore, based on m z (z), the mean and variance of w are estimated by LMMSE:

[0037]

[0038]

[0039] In the formula

[0040]

[0041]

[0042] Furthermore, calculate the message. The mean and variance are respectively:

[0043]

[0044]

[0045] In the formula, and

[0046] Compute message The mean and variance are respectively:

[0047]

[0048]

[0049]

[0050] In the formula, The normalization factor; the message m obtained X (X) is an estimate of the signal matrix X.

[0051] Furthermore, calculate the message. The mean and variance are respectively:

[0052]

[0053]

[0054] Compute message The mean and variance are respectively:

[0055]

[0056]

[0057]

[0058] In the formula, The normalization factor is used to obtain message m. s (S) is an estimate of the channel coefficient matrix S.

[0059] Furthermore, if the maximum number of iterations is exceeded, the process ends; otherwise, it is recalculated until the maximum number of iterations is exceeded.

[0060] Another object of the present invention is to provide a blind channel estimation and signal detection method for a large-scale MIMO system, the system comprising:

[0061] The probability modeling module is used to perform probability modeling on the channel coefficient matrix and signal matrix at the base station receiver to obtain the prior distributions p(S) and p(X) of S and X.

[0062] The initialization module is used to initialize the channel coefficient matrix based on the prior distribution described in Essence 1, thereby obtaining the initial values ​​of the channel coefficient matrix. and the covariance matrix V between its columns s Initial values ​​of the signal matrix and the covariance matrix U between its rows x ;

[0063] The matrix estimation module is used to estimate the channel coefficient matrix S and the signal matrix based on the received signal y using the GBF-HVMP algorithm;

[0064] The matrix estimation module is used to estimate the trace coefficient matrix obtained from pilot sequence elimination. With signal matrix estimate The permutation fuzziness and scale fuzziness are determined, and the results are obtained through calculation. The estimated value of the channel matrix is ​​obtained.

[0065] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0066] First, for existing blind channel estimation and signal detection methods for large-scale MIMO systems, we consider the design of the steering vector matrix and quantizer, and introduce a generalized matrix factorization algorithm based on hybrid vector message passing, achieving lower computational complexity and better channel estimation and signal detection performance. The details are as follows:

[0067] This invention models blind channel estimation and signal detection in large-scale MIMO systems as a generalized matrix factorization form. The proposed scheme considers the design of the steering vector matrix and quantizer, making it applicable to a wider range of large-scale MIMO systems.

[0068] This invention develops a generalized matrix factorization algorithm based on hybrid vector message passing for channel estimation and signal detection. Simulation results verify the superior performance of the proposed algorithm in channel estimation and signal detection, while achieving lower time complexity.

[0069] Second, considering the technical solution as a whole or from a product perspective, the technical effects and advantages of the technical solution to be protected by this invention are specifically described as follows:

[0070] This invention proposes a novel system model for blind channel estimation and signal detection in large-scale MIMO systems, taking into account the design of the steering vector matrix and quantizer, and can be applied to a wider range of scenarios. It introduces a generalized matrix factorization algorithm based on hybrid vector message passing for blind channel estimation and signal detection, achieving lower computational complexity and better channel estimation and signal detection performance.

[0071] This invention proposes a new system modeling and algorithm implementation for blind channel estimation and signal detection in large-scale MIMO systems. Compared with existing domestic and foreign systems and algorithms, it successfully achieves lower complexity and better blind channel estimation and signal detection performance.

[0072] Third, the blind channel estimation and signal detection method in large-scale MIMO systems provided by this invention has achieved significant technical progress in the following aspects:

[0073] 1) Refined Probabilistic Modeling: By performing detailed probabilistic modeling of the channel coefficient matrix and signal matrix, this method can more accurately estimate and process signals. This modeling approach helps to better understand and predict the behavior of signal transmission, thereby improving the performance of the entire system.

[0074] 2) Using the GBF-HVMP algorithm: The application of the GBF-HVMP algorithm outperforms traditional blind channel estimation methods in terms of computational efficiency and estimation accuracy. This algorithm improves the accuracy of channel estimation while reducing computational complexity.

[0075] 3) Eliminating permutation and scale ambiguities: Using pilot sequences to resolve permutation and scale ambiguities between the channel coefficient matrix estimates and the signal matrix estimates is a significant technological advancement. This helps improve the accuracy of the estimates and the overall system performance.

[0076] 4) Angle-domain channel representation: Dividing the signal's angle of arrival into multiple resolution intervals and establishing an angle-domain representation of the channel accordingly is an innovative processing method. It can more effectively utilize the sparsity of the channel, thereby improving the efficiency and accuracy of channel estimation.

[0077] 5) Signal processing considering quantization error: This method also considers the effect of element-by-element quantizer at the receiver due to precision limitations in quantized MIMO systems. This means that this method can better adapt to actual hardware limitations and improve the practical application performance of the system. Attached Figure Description

[0078] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0079] Figure 1 This is a system model diagram provided in an embodiment of the present invention;

[0080] Figure 2 This is a flowchart of the blind channel estimation and signal detection method for a large-scale MIMO system provided in this embodiment of the invention;

[0081] Figure 3 This is the GBF-HVMP algorithm factor graph provided in the embodiments of the present invention;

[0082] Figure 4 This is a structural diagram of a blind channel estimation and signal detection system for a large-scale MIMO system provided in an embodiment of the present invention;

[0083] Figure 5 This is a simulation result diagram of the channel and signal normalized mean square error (NMSE) provided in the embodiment of the present invention;

[0084] Figure 6 This is a simulation result diagram of the NMSE phase transition provided in the embodiment of the present invention;

[0085] Figure 7 This is a simulation result graph of the algorithm complexity (running time) provided in the embodiment of the present invention;

[0086] Figure 8 These are simulation results of different bit quantization methods provided in the embodiments of the present invention. Detailed Implementation

[0087] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0088] The present invention provides a blind channel estimation and signal detection method for large-scale MIMO systems, offering two specific embodiments and their detailed implementation schemes:

[0089] Example 1: Blind Channel Estimation and Signal Detection Based on Corner Domain Grid Partitioning

[0090] 1) System Configuration:

[0091] The base station is equipped with a uniform linear array of (N) receiving antennas.

[0092] It serves K single-antenna users.

[0093] The angular distribution of the incoming signal for each user is known.

[0094] 2) Corner domain grid partitioning:

[0095] The angle of arrival of the signal is divided into \(M\) resolution intervals, each interval being represented by \([\theta_{\text{min}},\theta_{\text{max}}]\).

[0096] Calculate each interval using the interval \(d\) and wavelength \(\lambda\).

[0097] 3) Blind channel estimation and signal detection algorithms:

[0098] Channel estimation is performed using the GBF-HVMP algorithm.

[0099] Channel sparsity modeling is performed using angle of arrival information.

[0100] Use probabilistic modeling and initialize the channel coefficient matrix and signal matrix from step one.

[0101] 4) Result Processing:

[0102] Permutation ambiguity and scale ambiguity are eliminated by pilot sequences.

[0103] Example 2: Mixed Signal Processing and Blind Channel Estimation

[0104] 1) System Configuration:

[0105] Same configuration as in Example 1.

[0106] 2) Signal preprocessing:

[0107] The received signal undergoes preliminary processing, including filtering and noise reduction.

[0108] 3) Blind channel estimation and signal detection:

[0109] Deep learning algorithms, such as convolutional neural networks (CNNs), are applied to perform blind channel estimation by taking advantage of the asymptotic orthogonality of large-scale MIMO channels.

[0110] The accuracy of the estimation is improved by combining the GBF-HVMP algorithm.

[0111] 4) Design of the guide vector matrix:

[0112] Design a guidance vector matrix that adapts to different angle-of-arrival ranges.

[0113] Determine the number of grid points and the number of bits in the quantizer.

[0114] 5) Signal recovery and optimization:

[0115] The estimated signal matrix is ​​post-processed, including the removal of residual ambiguity and noise.

[0116] Optimize the signal recovery process to improve overall performance.

[0117] Both examples are specific applications of blind channel estimation and signal detection in large-scale MIMO systems. The first example focuses on using angular domain information for blind channel estimation and signal detection, while the second example combines deep learning and traditional algorithms for more complex channel estimation and signal processing. Both approaches demonstrate advantages in improving channel estimation accuracy, reducing complexity, and enhancing system performance.

[0118] To address the problems existing in the prior art, this invention provides a blind channel estimation and signal detection method and system for large-scale MIMO systems.

[0119] This invention considers a massively MIMO-assisted uplink communication system, the system model of which is as follows: Figure 1 As shown, a base station is configured with a uniform linear array of N receiving antennas to serve K single-antenna users. Let θ k，i This represents the i-th angle of arrival of the incoming signal from the k-th user. It is the set of arrival angles.

[0120] The signal angle of arrival is divided into L resolution intervals, where the l-th interval is composed of... In this expression, d is the spacing between any two adjacent receiving antennas, λ is the propagation wavelength, and l∈{0,1,…,L-1}. Let… Covers angles from -90° to 90°. When When the channel between the k-th user and the base station in the angle domain can be represented as:

[0121]

[0122] In the formula These are sparse channel coefficients in the angular domain. It is a guide vector. We assume that the channel is block-fading with a coherence time of T, meaning that the channel remains unchanged within each transport block.

[0123] For each coherence time T, the signal of the k-th user is represented as: The set of all transmitted signals from the user terminal is The first N p The sequence is listed as a pilot sequence. The received signal can be given by the following formula.

[0124] y = φ(z),

[0125] z = Fw + n,

[0126]

[0127] w = vec(SX),

[0128] In the formula, φ(·) is the element-wise quantizer generated by the receiver in the quantization MIM0 system due to precision limitations. This represents the Kronecker product, vec(·) represents the vectorization operator, and z is the unquantized received signal. The channel coefficient matrix, Let n represent an AWGN vector. The elements of n are independent and identically distributed, and all follow a certain distribution. Distribution, where σ 2 This represents noise power.

[0129] like Figure 2 As shown in the figure, an embodiment of the present invention provides a blind channel estimation and signal detection method for a large-scale MIMO system; according to the parameter settings in Table 1, the specific steps of the simulation are as follows:

[0130] Table 1 Main Simulation Parameters

[0131]

[0132] S1, at the base station receiver, performs probabilistic modeling of the prior information of the channel and signal. For the prior distributions of the channel coefficient matrix S and the signal matrix X, it is assumed that the elements of X are independent and identically distributed, and generated from a complex Gaussian distribution, i.e.

[0133]

[0134] In the formula Let S be the energy of the signal. Since large-scale MIMO channels in the angular domain are sparse, the prior distribution of the channel coefficient matrix S is modeled as a product of Bernoulli-Gaussian distributions:

[0135]

[0136] In the formula, ρ is the sparsity. It is the energy of the channel coefficient.

[0137] S2 is initialized based on the prior distribution in S1 to obtain the initial values ​​of the channel coefficient matrix. and the covariance matrix V between its columns S Initial values ​​of the signal matrix and the covariance matrix U between its rows X .

[0138] S3 uses the GBF-HVMP algorithm to estimate the channel coefficient matrix S and the signal matrix based on the received signal y. The factor graph of the GBF-HVMP algorithm is shown below. Figure 3 As shown, the message is calculated first. Its mean is

[0139]

[0140] Compute message Its mean and variance are respectively

[0141]

[0142]

[0143] In the formula, is the normalization factor. The obtained message mz(z) is an estimate of the unquantized received signal.

[0144] Based on mz(z), w is estimated using LMMSE, with its mean and variance being respectively...

[0145]

[0146]

[0147] In the formula

[0148]

[0149]

[0150] Compute message Its mean and variance are respectively

[0151]

[0152]

[0153] In the formula and

[0154] Compute message Its mean and variance are respectively

[0155]

[0156]

[0157]

[0158] In the formula This is the normalization factor. The received message m X (X) is an estimate of the signal matrix X.

[0159] Compute message Its mean and variance are respectively

[0160]

[0161]

[0162] Compute message Its mean and variance are respectively

[0163]

[0164]

[0165]

[0166] In the formula This is the normalization factor. The received message m s (S) is an estimate of the channel coefficient matrix S.

[0167] If the maximum number of iterations is exceeded, the process ends; otherwise, proceed to S3.

[0168] S4, Estimated value of the trace coefficient matrix obtained from pilot sequence elimination. With signal matrix estimate The permutation fuzziness and scale fuzziness are determined, and the results are obtained through calculation. The estimated value of the channel moment is obtained.

[0169] like Figure 4 As shown, this embodiment of the invention provides a blind channel estimation and signal detection system for a large-scale MIMO system. The system includes:

[0170] The probability modeling module is used to perform probability modeling on the channel coefficient matrix and signal matrix at the base station receiver to obtain the prior distributions p(S) and p(X) of S and X.

[0171] The initialization module is used to initialize the channel coefficient matrix based on the prior distribution described in Essence 1, thereby obtaining the initial values ​​of the channel coefficient matrix. and the covariance matrix V between its columns S Initial values ​​of the signal matrix and the covariance matrix U between its rows X ;

[0172] The matrix estimation module is used to estimate the channel coefficient matrix S and the signal matrix X based on the received signal y using the GBF-HVMP algorithm;

[0173] The matrix estimation module is used to estimate the trace coefficient matrix obtained from pilot sequence elimination. With signal matrix estimate The permutation fuzziness and scale fuzziness are determined, and the results are obtained through calculation. The estimated value of the channel matrix is ​​obtained.

[0174] For blind channel estimation and signal detection scenarios in specific large-scale MIMO systems, considering different parameters such as the number of grid points, signal-to-noise ratio, and number of users, the specific results are as follows.

[0175] Figure 5 The NMSE of signal X and channel H at different SNR levels is shown. Over the entire considered SNR range, the GBF-HVMP algorithm exhibits a significant NMSE gain of over 10 dB compared to the baseline method. Figure 6 The algorithm's NMSE phase transition is illustrated, describing the NMSE of signals X with different sparsity ρ and user numbers K. The SNR is fixed at 20 dB. The top four subplots correspond to the case where L = 64, and the bottom four subplots correspond to the case where L = 256; other parameters are consistent with Table 1. In both cases, the proposed GBF-HVMP algorithm exhibits better phase transition curves than the baseline.

[0176] exist Figure 7In the evaluation, the computational complexity of the algorithm is assessed based on the running time, where the SNR is fixed at 20dB and other parameters are consistent with those in Table 1. The algorithm's stopping criterion is defined as X's NMSE reaching -20dB. Figure 7 The curves of the baseline algorithms are incomplete because the NMSE of the corresponding algorithms cannot reach -20dB when the number of users is high. Compared with all baseline algorithms, the proposed GBF-HVMP algorithm achieves a faster convergence speed and is significantly better than all baselines, which clearly demonstrates the superiority of the proposed solution.

[0177] exist Figure 8 In this paper, we consider quantized MIMO systems with different bit counts. We set the number of users to K = 20, and other parameters are the same as in Table 1. Figure 8 This paper illustrates the impact of 1-3 bit quantization on system performance. Under low SNR conditions, the performance degradation of 3-bit quantization is negligible; however, for 1-bit and 2-bit quantization, the performance degradation is considerable. As the signal-to-noise ratio (SNR) increases, the performance degradation of all quantization systems becomes more significant. This is because the benefits of a high SNR are limited by the quantization process. Other algorithms cannot be used in this scenario, demonstrating the superiority of the proposed solution.

[0178] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A blind channel estimation and signal detection method for a large-scale MIMO system, characterized in that, Includes the following steps: Step 1: At the base station receiver, adjust the channel coefficient matrix. With signal matrix Probability modeling is performed to obtain the prior distribution. and Set the guiding vector matrix Number of corner grid points And based on the number of bits of the quantizer Configure element-wise quantizer ; Step 2: Initialize according to the prior distribution to obtain the initial values ​​of the channel coefficient matrix. and its column covariance matrix and the initial values ​​of the signal matrix and its row covariance matrix ; Step 3: Use the GBF-HVMP algorithm based on the received signal. Iterative estimation of channel coefficient matrix and signal matrix The GBF-HVMP algorithm is an algorithm that uses Hybrid Vector Message Passing (HVMP) to solve the Generalized Bilinear Factorization (GBF) problem, including: (1) Using the channel coefficient matrix estimate of the current iteration and signal matrix estimates Construct matrix product Calculate the predicted mean of the unquantized received signal. Based on the number of bits Element-wise quantizer with defined quantization rules Received signal and noise power, update the posterior mean and variance of the unquantized received signal. ; (2) Based on and The product variable is updated by estimating the linear minimum mean square error. The estimated value and variance ; (3) According to , And the current and Calculate the pointing signal matrix The complex matrix Gaussian message, and the message is compared with the prior distribution. Combine and update the signal matrix estimate. and its row covariance matrix ; (4) According to , And the updated and Calculate the pointing channel coefficient matrix The complex matrix Gaussian message, and the message is compared with the prior distribution. Combine and update the estimated values ​​of the channel coefficient matrix. and its column covariance matrix ; (5) Determine whether the current iteration round has reached the preset maximum iteration round; if not, return to step (1); if it has, output the maximum iteration round. and ; Step 4: Eliminate the estimated channel coefficient matrix based on the pilot sequence. With signal matrix estimate The permutation fuzziness and scale fuzziness are determined, and the results are obtained through calculation. Obtain the estimated value of the channel matrix. .

2. The blind channel estimation and signal detection method for a large-scale MIMO system as described in claim 1, characterized in that, Base station deployment is set A uniform linear array of root receiving antennas serves For a single-antenna user, Indicates from the The first user's incoming signal An angle of arrival, It is the set of arrival angles.

3. The blind channel estimation and signal detection method for a large-scale MIMO system as described in claim 2, characterized in that, The signal arrival angle is divided into... The resolution range, of which the first The intervals are composed of It means that in the formula It is the interval between any two adjacent receiving antennas. It is the propagation wavelength. ;make Coverage from arrive From the angle, when At that time, the first in the angle domain The channel between a user and a base station is represented as follows: In the formula, These are sparse channel coefficients in the angular domain. It is a guide vector. The channel is block fading, and the coherence time is... The channel remains unchanged within each transport block.

4. The blind channel estimation and signal detection method for a large-scale MIMO system as described in claim 3, characterized in that, For each coherence time , No. The signal of each user is represented as The set of all transmitted signals from the user terminal is ,forward The sequence is listed as a pilot sequence; the received signal is given by the following formula: In the formula, For the number of bits Configuring an element-wise quantizer, Represents the Kronecker product. This indicates vectorized operators. The received signal is unquantized. The channel coefficient matrix, It is an additive white Gaussian noise vector; Each element is independently and identically distributed and follows a set pattern. ,in This represents noise power.

5. The blind channel estimation and signal detection method for a large-scale MIMO system as described in claim 4, characterized in that, Step 3: First, calculate the message. The mean is: Compute message The mean and variance are as follows: In the formula, The normalization factor; the message received This is an estimate of the unquantized received signal.

6. The blind channel estimation and signal detection method for a large-scale MIMO system as described in claim 5, characterized in that, based on Obtained through LMMSE The estimates, mean, and variance are as follows: In the formula 。 7. The blind channel estimation and signal detection method for a large-scale MIMO system as described in claim 6, characterized in that, Compute message The mean and variance are as follows: In the formula, and ; Compute message The mean and variance are as follows: In the formula, The normalization factor; the message received For the signal matrix The estimate.

8. The blind channel estimation and signal detection method for a large-scale MIMO system as described in claim 7, characterized in that, Compute message The mean and variance are as follows: Compute message The mean and variance are as follows: In the formula, The message obtained is the normalization factor. For the channel coefficient matrix The estimate.

9. The blind channel estimation and signal detection method for a large-scale MIMO system as described in claim 8, characterized in that, If the maximum number of iterations is exceeded, the process ends; otherwise, the calculation is recalculated until the maximum number of iterations is exceeded.

10. A blind channel estimation and signal detection system for a large-scale MIMO system as described in any one of claims 1 to 9, characterized in that, The system includes: The probabilistic modeling module is used to perform probabilistic modeling of the channel coefficient matrix and signal matrix at the base station receiver, obtaining... and prior distribution and ; The initialization module is used to initialize the channel coefficient matrix based on the prior distribution in step one, thereby obtaining the initial values ​​of the channel coefficient matrix. and its column covariance matrix and the initial values ​​of the signal matrix and its row covariance matrix ; The matrix estimation module is used to perform the GBF-HVMP algorithm based on the received signal. Estimating the channel coefficient matrix and signal matrix; The matrix estimation module is used to estimate the channel coefficient matrix obtained from pilot sequence elimination. With signal matrix estimate The permutation fuzziness and scale fuzziness are determined, and the results are obtained through calculation. The estimated value of the channel matrix is ​​obtained.

Images