Hybrid distributed signal detection apparatus and method for cell-free massive MIMO systems

Abstract

The application discloses a kind of hybrid distributed signal detection device and method of de-cell large-scale MIMO system, the device includes: signal receiving module is used to receive sending signal;Radio frequency processing module is used to carry out radio frequency processing to receiving signal;Channel estimation module is used to obtain the channel estimation information between each AP and user;Signal detection module is used to adopt MRC method or EP method to the user served by AP carries out signal detection, obtains partial signal detection result;Information merging processing module is used to carry out merging processing according to the partial signal detection result of each AP, obtains the global signal detection result of each user.The application effectively reduces the data quantity of front link between AP and CPU and the calculation complexity of CPU, combines the complexity advantage of MRC method and the performance advantage of EP method, while guaranteeing distributed detection performance, maintains low calculation complexity, meets the implementation needs of system, has significant application value.

Description

Technical Field

[0001] This invention belongs to the field of signal processing technology in wireless communication, and particularly relates to a hybrid distributed signal detection device and method for decellularized massive MIMO systems. Background Technology

[0002] Cellular massive multiple-input multiple-output (MIMO) networks distribute a large number of access points (APs) over a wide area and connect them to a central processing unit (CPU) via fronthaul links. Because the distance between APs and users is reduced, path loss is effectively decreased. Simultaneously, the coordinated transmission of multiple APs converts interference signals into effective signals, significantly improving access performance and coverage for users at the cell edge. This enables 6G mobile communication systems to provide ultra-high throughput and ultra-high system capacity wireless access and transmission services.

[0003] The application of decellularized massive MIMO technology faces numerous challenges, one of which is the implementation of low-complexity detectors with near-optimal maximum likelihood (ML) detection performance. Currently, most decellularized massive MIMO systems are based on centralized signal detection, where each access point (AP) transmits the received signal to the CPU via a fronthaul link for centralized signal processing. With the significant increase in user terminals and signal bandwidth, the fronthaul data volume exceeds existing high-speed interconnect bandwidth standards, and the ultra-dense deployment of APs brings extremely high computational complexity to the CPU. To address this, a distributed signal detection scheme is proposed, where each AP performs independent, parallel signal detection, and the detection results are merged at the CPU, thereby reducing the complexity of centralized CPU processing and the fronthaul data volume. However, most existing detection schemes employ simple linear algorithms, which increase interference between users when an AP serves multiple users simultaneously. Furthermore, the lack of global information sharing among APs leads to a significant performance loss in distributed signal detection compared to centralized detection. Furthermore, existing distributed signal detection schemes are all based on the assumption that each user is served by all APs, without considering a flexibly scalable dynamic cooperative cluster structure. In real-world scenarios, some APs are far from users, and the performance gain provided by these APs is limited while introducing additional interference. Moreover, this structure makes decellularized networks difficult to scale flexibly and incurs significant computational burden and power consumption. To address this, a user-centric dynamic cooperative cluster structure is proposed, where each user is only served by a subset of APs with better channel quality. Due to the random distribution of users, the number of users served by each AP varies considerably. For APs serving a large number of users, using a simple linear detector cannot fully utilize the diversity gain among multiple users, resulting in performance loss. Conversely, for APs serving a small number of users, using a complex nonlinear detector provides limited performance gains but incurs a significant computational burden.

[0004] Therefore, to address the above issues, it is necessary to design a new hybrid distributed signal detection device and method for decellularized massive MIMO systems. This method should flexibly process signals based on the actual overload conditions of each access point (AP) to adapt to the characteristics and requirements of actual decellularized massive MIMO systems and achieve the best system performance-complexity balance. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a hybrid distributed signal detection device and method for decellularized massive MIMO systems. This invention enables reliable distributed signal detection with low computational complexity, meeting the implementation requirements of practical decellularized massive MIMO systems.

[0006] The objective of this invention is achieved through the following technical solution: A first aspect of this invention provides a hybrid distributed signal detection device for a decellularized massive MIMO system, comprising:

[0007] The signal receiving module is used to receive transmitted signals from the served users for the activated access point; wherein the transmitted signals include pilot signals and data signals;

[0008] The radio frequency (RF) processing module is used to perform RF processing on the signals received by the signal receiving module; the RF processing includes filtering, low-noise amplification, and demodulation.

[0009] The channel estimation module is used to estimate the channel information between each access point and the communication user based on the pilot signals sent by the user and using the minimum mean square error estimation method.

[0010] The signal detection module, based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, uses the maximum ratio combining method or the expectation value propagation method to perform signal detection on the users served by the access point, in order to obtain local signal detection results; and

[0011] The information merging and processing module is used to merge the local signal detection results from each access point using a hybrid linear-nonlinear information fusion rule to calculate the global signal detection result for each user.

[0012] Furthermore, the pilot signal received by the signal receiving module is represented as follows:

[0013]

[0014] Among them, Z m This represents the received pilot signal at the m-th access point. This represents the channel vector between the m-th access point and the k-th user. This indicates that the expression follows a pattern with a mean of 0 and a covariance of R. mk The complex Gaussian distribution, This represents the pilot signal assigned to the k-th user, and ||c k || 2 =τ p ,(·) T τ represents the transpose operation of a vector or matrix. p Indicates the pilot signal transmission time slot. Represents N×τ p A Vaticaned white Gaussian noise matrix, wherein each element follows a mean of 0 and a variance of σ. 2 The complex Gaussian distribution is given by N, which represents the number of antennas at each access point.

[0015] The channel estimation module uses the minimum mean square error estimation method to estimate the channel information between each access point and the communication user. The calculation formula is as follows:

[0016]

[0017] in, h mk The estimated value, I N represents an N-dimensional identity matrix, and (·)* represents the conjugate operation of a vector or matrix.

[0018] A second aspect of this invention provides a signal detection method based on the above-described hybrid distributed signal detection device for decellularized massive MIMO systems, characterized by comprising the following steps:

[0019] (1) Construct an uplink decellularized large-scale MIMO dynamic cooperative cluster communication model based on the user's transmitted signal and the channel estimation information obtained by the channel estimation module;

[0020] (2) Calculate the overload ratio of each activated access point based on the number of antennas configured for each access point and the set of users served by that access point;

[0021] (3) Multiple activated access points perform independent and parallel distributed signal detection: For each activated access point, an overload ratio threshold is preset for each activated access point. It is determined whether the calculated overload ratio is less than the preset overload threshold. If the calculated overload ratio is less than the preset overload threshold, the maximum ratio combining method is used to perform signal detection on the users served by the activated access point based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, so as to obtain the local signal detection result; otherwise, the expectation value propagation method is used to perform signal detection on the users served by the activated access point based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, so as to obtain the local signal detection result.

[0022] (4) The signal detection module of each access point transmits the local detection results obtained in step (3) to the information merging and processing module of the centralized processing unit through the fronthaul link, and merges them using a hybrid linear-nonlinear information fusion rule to obtain the global signal detection results of each user.

[0023] Furthermore, step (1) specifically includes:

[0024] The service identifier matrix D of the k-th user and the m-th access point mk Represented as:

[0025]

[0026] Among them, IN Represents an N-dimensional identity matrix, 0 N×N Represents an N×N dimensional zero matrix. This represents the set of access points that provide services to the k-th user.

[0027] The set of users served by the m-th access point The expression is:

[0028]

[0029] in, The size is tr(·) represents the trace operation of a matrix;

[0030] The expression for the communication model is:

[0031]

[0032] Among them, y m This represents the received signal of the m-th access point. This represents the set of active access points. For N×K m 3D channel matrix, [·] T This represents the transpose operation of a [·] matrix. For K m Each user transmits data signals, and each element of these signals is a constellation point in the constellation set Ω. It is an additive white Gaussian noise vector. This indicates that the expression follows a pattern with a mean of 0 and a covariance of σ. 2 I N The complex Gaussian distribution.

[0033] Furthermore, the formula for calculating the overload ratio is:

[0034] α m =K m / N (6)

[0035] Where, α m K represents the overload ratio of the m-th activated access point, N is the number of antennas configured for the m-th activated access point, and K is the number of antennas configured for the m-th activated access point. m The size of the set of users served by the m-th activated access point.

[0036] Further, in step (3), the use of the maximum ratio combining method to perform signal detection on the users served by the activated access point based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, in order to obtain local signal detection results, specifically includes:

[0037] Local signal detection result x mRepresented as:

[0038]

[0039] in, Let represent the Gram matrix, (·) H The `diag` operator represents the conjugate transpose of a matrix, and `diag(·)` extracts the diagonal elements of the matrix. Represents the maximum ratio merge vector;

[0040] Estimation error caused by using the maximum ratio combining method for signal detection Represented as:

[0041]

[0042] in, K represents m An identity matrix of dimension (·) -1 The expression represents the matrix inversion operation, and diag(diag(·)) represents a diagonal matrix formed by the diagonal elements of the matrix (·).

[0043] Further, in step (3), the use of the expectation value propagation method to perform signal detection on the users served by the activated access point based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, in order to obtain local signal detection results, specifically includes:

[0044] The joint distribution of the transmitted and received signals at the m-th access point, p(x) m ,y m Decomposed into:

[0045]

[0046] Where p(y) m |x m p(x) represents the conditional distribution. m ) represents x m The prior distribution of y m,n Represents vector y m The nth element, x m,k Represents vector x m The kth element;

[0047] make and They represent the variable nodes x in the t-th iteration, respectively. k Passed to factor node f n The mean and variance, and These represent the factors from node f. n Passed to variable node x kThe mean and variance are calculated using the following formulas:

[0048]

[0049]

[0050] Among them, h m,n,k and h m,n,k' Representing the channel vector h respectively mk The kth and k'th elements, and Representing sets The posterior mean and posterior variance of the k-th user are expressed as:

[0051]

[0052] in, p represents the expectation operation. (t) (x m,k |y m ) represents the posterior probability distribution of the constellation point for the k-th user, and its calculation formula is:

[0053]

[0054] Where ∝ represents the normalization operation, a∈Ω represents a constellation point, and p (t) (x m,k =a) represents the prior probability of a constellation point, initialized to p. (t) (x m,k =a) = 1 / Q, where Q represents the size of the constellation set;

[0055] All factor nodes f n To variable node x k cumulative mean and cumulative variance They are represented as follows:

[0056]

[0057] in, and These represent the factors from node f in the t-th iteration, respectively. n Passed to variable node x k The mean and variance, factor nodes f n Represents the likelihood function, variable node x k Represents the user.

[0058] Further, in step (4), the merging of signals using a hybrid linear-nonlinear information fusion rule to obtain the global signal detection results for each user specifically includes:

[0059] For the k-th user, define and Let each be a set of access points serving the k-th user using the maximum ratio merging method and the expectation value propagation method. The local detection results of multiple access points obtained using the maximum ratio merging method are merged using a linear mixing method to obtain a first merged detection result. The local detection results of multiple access points obtained using the expectation value propagation method are merged using a nonlinear mixing method to obtain a second merged detection result. The first merged detection result and the second merged detection result are merged using a linear mixing method to obtain the global signal detection result for the k-th user.

[0060] Furthermore, the method for obtaining the first merged detection result specifically includes:

[0061] The local detection results of the m-th access point for the k-th user obtained using the maximum ratio merging method Represented as:

[0062]

[0063] in, Indicates with x m,k Independent estimation error terms, with a mean of 0 and a variance of . The complex Gaussian distribution;

[0064] The first merged detection result is obtained by linearly weighting the local detection results of multiple access points obtained by the maximum ratio merging method for the k-th user. The expression is as follows:

[0065]

[0066]

[0067] in, express The corresponding weighting factors satisfy

[0068] Furthermore, the method for obtaining the second merged detection result specifically includes:

[0069] After reaching the preset maximum number of iterations T, based on the local detection results of the k-th user obtained using the expectation propagation method, the cumulative mean is... and cumulative variance The data is transmitted to the centralized processing unit via the fronthaul link for nonlinear merging, resulting in:

[0070]

[0071]

[0072] in, Let Variance be the variance of the k-th user. The mean of the k-th user;

[0073] The posterior probability distribution of the k-th user is represented as follows:

[0074]

[0075] The posterior mean and posterior variance are expressed as follows:

[0076]

[0077] in, Let represent the posterior mean of the k-th user, which is the second merged detection result for the k-th user. Let represent the posterior variance of the k-th user.

[0078] Furthermore, the method for obtaining the user's global signal detection results specifically includes:

[0079] A linear merging calculation is performed based on the first and second merged detection results to obtain the user's global signal detection result, expressed as follows:

[0080]

[0081] in, and These represent the weighting factors for signal detection using the maximum ratio combining method and the expectation value propagation method, respectively, and their calculation formulas are as follows:

[0082]

[0083]

[0084] in, and The calculation formulas are as follows:

[0085]

[0086]

[0087] in, This represents the number of access points that use the expectation value propagation method for signal detection. This represents the average number of users who used the expected value propagation method to detect the signal.

[0088] The beneficial effects of this invention are as follows: Firstly, it employs hybrid distributed signal detection, which effectively reduces the overhead of the fronthaul link between the AP and the CPU, as well as the computational load on the CPU, thus significantly reducing the amount of data transmitted through the fronthaul link. Secondly, considering the random distribution characteristics of users in decellularized massive MIMO systems, this invention adopts a user-centric dynamic cooperative cluster architecture. Each AP only provides services to users with good channel quality, effectively reducing interference between users within the group and the computational complexity of AP distributed signal detection. Thirdly, based on the actual overload ratio of the AP, this invention employs a flexible signal detection method, combining the complexity advantages of the MRC method with the performance advantages of the EP method. This ensures low computational complexity while maintaining distributed detection performance, meeting the implementation requirements of decellularized massive MIMO systems and demonstrating significant application value. Fourthly, based on the hybrid information fusion rules, this invention designs a corresponding merging scheme at the CPU based on the characteristics of the distributed detection algorithm. For the EP method, it proposes an efficient nonlinear merging scheme, enhancing information fusion between APs and improving the performance of distributed detection. Attached Figure Description

[0089] Figure 1 This is a schematic diagram of the user-centric decellularized massive MIMO system of the present invention;

[0090] Figure 2 This is a structural block diagram of the hybrid distributed signal detection device for the decellularized massive MIMO system of the present invention;

[0091] Figure 3 This is a flowchart of the hybrid distributed signal detection (MRC-EP) method for decellularized massive MIMO systems of the present invention;

[0092] Figure 4 This is a flowchart of the hybrid distributed signal detection method of the present invention;

[0093] Figure 5 This is the joint distribution factor diagram in this invention;

[0094] Figure 6 This is a comparison chart of the bit error rates of the signal detection method of the present invention and existing signal detection methods;

[0095] Figure 7 This is a comparison chart of the computational complexity of the signal detection method of the present invention and existing signal detection methods. Detailed Implementation

[0096] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0097] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0098] It should be understood that although the terms first, second, third, etc., may be used in this invention to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first information may also be referred to as second information without departing from the scope of this invention, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0099] The present invention will now be described in detail with reference to the accompanying drawings. Unless otherwise specified, the features of the following embodiments and implementations can be combined with each other.

[0100] In this embodiment of the invention, a user-centric uplink decellularized massive MIMO system is considered, such as... Figure 1 As shown. Within a certain area, there are M nodes with a spacing of L. A There are uniformly distributed access points (APs), each with N antennas connected to the CPU via a fronthaul link. There are K randomly distributed single-antenna users, each of whom interacts with APs within a radius R centered on them.

[0101] See Figure 2 The hybrid distributed signal detection device for a decellularized massive MIMO system of the present invention includes a signal receiving module, a radio frequency processing module, a channel estimation module, a signal detection module, and an information merging processing module. The signal receiving module, radio frequency processing module, channel estimation module, and signal detection module are signal processing devices for each AP, and the information merging processing module is a signal processing device for the CPU.

[0102] In this embodiment, the signal receiving module is used to receive transmitted signals from the served users for the activated AP; wherein the transmitted signals include pilot signals and data signals.

[0103] It should be understood that being activated means that the AP is providing service to the user. The received signal is the signal obtained at the base station after the pilot signal and data signal have passed through the channel and noise superimposed. The pilot signal is information known at the base station.

[0104] In this embodiment, the radio frequency processing module is used to perform radio frequency processing on the signal received by the signal receiving module; wherein, the radio frequency processing includes filtering, low noise amplification and demodulation processing.

[0105] It should be understood that the received signal is processed using existing standard methods.

[0106] In this embodiment, the channel estimation module is used to estimate the channel information between each AP and the communication user based on the pilot signal sent by the user and using the minimum mean square error estimation method.

[0107] Specifically, consider τ c The coherence interval of the time slot, where τ p Time slots are used for pilot signal transmission, τ d Time slots are used for data signal transmission, and have τ p Two mutually orthogonal pilot signals Used for channel estimation. Assume K ≤ τ p And the pilot signal allocated to the k-th user (k = 1, 2, ..., K) is And ||c k || 2 =τ p ,in Represents the space of complex numbers. c k τ is a complex space p 1-dimensional vector, ||·|| 2 Let the square operation of the absolute value of a vector be represented. Then, the received pilot signal Z at the m-th AP (m = 1, 2, ..., M) is... m Represented as:

[0108]

[0109] in, This represents the channel vector between the m-th AP and the k-th user. This indicates that the expression follows a pattern with a mean of 0 and a covariance of R. mk The complex Gaussian distribution of (·) T The transpose operation represents a vector or matrix. Represents N×τ pA Vaticaned white Gaussian noise matrix, wherein each element follows a mean of 0 and a variance of σ. 2 The complex Gaussian distribution. The formula for channel estimation based on the minimum mean square error estimation method is expressed as:

[0110]

[0111] in, h mk The estimated value, I N Let represent an N×N dimensional identity matrix, (·) * Represents the conjugate operation of vectors or matrices.

[0112] It should be noted that this invention aims to solve the user-centric distributed signal detection problem of decellularized massive MIMO, therefore it assumes that the channel estimation is an ideal estimation, i.e.

[0113] In this embodiment, the signal detection module is used to perform signal detection on the users served by the AP based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, using the maximum ratio combining (MRC) method or the expectation propagation (EP) method to obtain local signal detection results.

[0114] It should be understood that the overload ratio of the AP can be calculated based on the channel estimation information. Then, based on the overload ratio of the AP, either the low-complexity MRC method or the high-performance EP method can be selected to perform signal detection on the users served by the AP.

[0115] In this embodiment, the information merging processing module is used to perform merging processing based on the local signal detection results from each AP using a hybrid linear-nonlinear information fusion rule, and calculate the global signal detection result for each user.

[0116] It should be understood that after the CPU's information merging and processing module receives the local signal detection results from each AP, it adopts the corresponding linear or nonlinear information fusion rules based on the characteristics of the MRC method or EP method, and performs linear merging processing on the fusion results to calculate the global signal detection results for each user.

[0117] It is worth mentioning that the embodiments of the present invention also provide a hybrid distributed signal detection method for decellularized massive MIMO systems, which is implemented based on the hybrid distributed signal detection device for decellularized massive MIMO systems in the above embodiments.

[0118] See Figure 3The hybrid distributed signal detection (MRC-EP) method for decellularized massive MIMO systems specifically includes the following steps:

[0119] (1) Construct an uplink decellularized large-scale MIMO dynamic cooperative cluster communication model based on the user's transmitted signal and the channel estimation information obtained by the channel estimation module.

[0120] Specifically, let Let D represent the set of APs providing services to the k-th user. Then, the service identifier matrix D between the k-th user and the m-th AP is... mk It can be represented as:

[0121]

[0122] Among them, I N Represents an N-dimensional identity matrix, 0 N×N Let represent an N×N dimensional zero matrix. Accordingly, define . Let m be the set of users served by the m-th AP, and its expression is:

[0123]

[0124] in, The size is tr(·) represents the trace operation of a matrix.

[0125] Channel estimation information h obtained from the channel estimation module mk The received signal y of the m-th AP m Represented as:

[0126]

[0127] in, This represents the set of APs that provide services to users (are activated). For N×K m 3D channel matrix, [·] T This represents the transpose operation of a [·] matrix. For K m Each user transmits data signals, and each element of these signals is a constellation point in the constellation set Ω. It is an additive white Gaussian noise vector. This indicates that the expression follows a pattern with a mean of 0 and a covariance of σ. 2 I N The complex Gaussian distribution.

[0128] (2) Calculate the overload ratio of each activated AP based on the number of antennas configured for each AP and the set of users served by the AP.

[0129] For each activated AP, the overload ratio is calculated using the following formula:

[0130] α m =K m / N (6)

[0131] Where, α m K represents the overload ratio of the m-th activated AP, N is the number of antennas configured in the m-th activated AP, and K m The size of the set of users served by the m-th activated AP.

[0132] (3) Multiple activated APs perform independent and parallel distributed signal detection: For each activated AP, an overload ratio threshold γ is preset for each activated AP, and the calculated overload ratio α is judged. m Is it less than the preset overload threshold γ? If the calculated overload ratio α m If the signal is less than the preset overload threshold γ, the low-computational-complexity MRC method is used to detect the signal of the user served by the activated AP based on the data signal processed by the RF processing module and the channel estimation information obtained by the channel estimation module, in order to obtain local signal detection results; otherwise, the high-performance nonlinear EP method is used to detect the signal of the user served by the activated AP based on the data signal processed by the RF processing module and the channel estimation information obtained by the channel estimation module, in order to obtain local signal detection results, such as... Figure 4 As shown below, the signal detection process will be explained in detail using the m-th AP as an example.

[0133] It should be understood that signal detection is the process of recovering the signal sent by the user from the receiving end.

[0134] ①When α m When γ < γ, a low-computational-complexity MRC signal detection method is used, and its local signal detection result x m Represented as:

[0135]

[0136] in, Let represent the Gram matrix, (·) H The `diag` operator represents the conjugate transpose of a matrix, and `diag(·)` extracts the diagonal elements of the matrix. This represents the MRC vector. The estimation error introduced by using the MRC method for signal detection... It can be represented as:

[0137]

[0138] in, K represents m An identity matrix of dimension (·) -1The expression represents the matrix inversion operation, and diag(diag(·)) represents a diagonal matrix formed by the diagonal elements of the matrix (·).

[0139] ②When α m When ≥γ, a high-performance nonlinear EP signal detection method is adopted. The joint distribution p(x) of the m-th AP transmit and receive signals. m ,y m It can be broken down into:

[0140]

[0141] Where p(y) m |x m p(x) represents the conditional distribution. m ) represents x m The prior distribution of y m,n Represents vector y m The nth element, x m,k Represents vector x m The k-th element. Formula (9) can be represented by a factor graph, such as Figure 5 As shown, there is K. m There are N variable nodes (VN), N factor nodes (FN), and K... m Each of the prior nodes (PN) represents the user, the receiver likelihood function, and prior information, respectively. The EP method obtains the posterior distribution of the user signal by iteratively passing the mean and variance among interconnected variable and factor nodes. Let... and They represent the variable nodes x in the t-th iteration, respectively. k Passed to factor node f n The mean and variance, and These represent the factors from node f. n Passed to variable node x k The mean and variance are calculated using the following formulas:

[0142]

[0143]

[0144] Among them, h m,n,k and h m,n,k' Representing the channel vector h respectively mk The kth and k'th elements, and Representing sets The posterior mean and posterior variance of the k-th user are expressed as:

[0145]

[0146] in, p represents the expectation operation. (t) (x m,k |y m ) represents the posterior probability distribution of the constellation point for the k-th user, which is calculated as follows:

[0147]

[0148] Where ∝ represents the normalization operation, a∈Ω represents a constellation point, and p (t) (x m,k =a) represents the prior probability of a constellation point, initialized to p. (t) (x m,k =a) = 1 / Q, where Q represents the size of the constellation set. and These represent the points from all factor nodes to the variable node x. k The cumulative mean and cumulative variance are expressed as:

[0149]

[0150] It should be understood that the cumulative mean and cumulative variance are the local detection results obtained by using the EP method for signal detection.

[0151] (4) The signal detection module of each AP transmits the local detection results obtained in step (3) to the information merging and processing module of the CPU through the fronthaul link. The merged results are then obtained using a hybrid linear-nonlinear information fusion rule to acquire the global signal detection results for each user. Figure 4 As shown.

[0152] It should be understood that after each active AP completes independent signal detection, it transmits the local detection results to the CPU through the fronthaul link, and the CPU merges the local signal detection results from each AP.

[0153] Specifically, for the k-th user, define and Given the sets of APs serving the k-th user using the MRC method and the EP method respectively, the local detection results of multiple APs obtained by the MRC method are linearly merged using a linear mixing method to obtain the first merged detection result; the local detection results of multiple APs obtained by the EP method are nonlinearly merged using a nonlinear mixing method to obtain the second merged detection result; the first merged detection result and the second merged detection result are linearly merged using a linear mixing method to obtain the global signal detection result for the k-th user.

[0154] ①When At that time, the local detection result of the k-th user and the m-th AP. It can be represented as:

[0155]

[0156] in, Indicates with x m,k Independent estimation error terms, with a mean of 0 and a variance of . If the APs detected by MRC follow a complex Gaussian distribution, then the linear weighted merging of all APs can be expressed as follows:

[0157]

[0158] in, express The corresponding weighting factors satisfy Based on the signal-to-interference-plus-noise ratio (SINR) maximization criterion and using the Lagrange multiplier method, the optimal weighting factor can be calculated as follows:

[0159]

[0160] Among them, error variance It is given by formula (8).

[0161] ②When To address the message-passing characteristics of the EP algorithm, this invention proposes a nonlinear fusion scheme. After reaching a preset maximum number of iterations T, the AP detected using the EP method will accumulate an average value. and cumulative variance The data is passed to the CPU via the fronthaul link for merging, resulting in:

[0162]

[0163]

[0164] in, Let Variance be the variance of the k-th user. Let be the mean of the k-th user.

[0165] Furthermore, the posterior probability distribution of the k-th user can be obtained, expressed as:

[0166]

[0167] The posterior mean and posterior variance can be expressed as:

[0168]

[0169] in, Let represent the posterior mean of the k-th user, which is the second merged detection result for the k-th user. Let represent the posterior variance of the k-th user.

[0170] According to formulas (18)-(21), the set comes from the T-th iteration. All nodes with variable x k The posterior probability of a user is obtained by jointly calculating the connected factor nodes, which is equivalent to a centralized calculation method. Therefore, the proposed scheme is superior to the simple linear weighted summation and merging method.

[0171] ③ After the MRC and EP distributed local detection results are merged, the first merged detection result and the second merged detection result are then linearly weighted and merged to obtain the user's global signal detection result, which can be expressed as:

[0172]

[0173] in, and The weighting factors for signal detection using the MRC and EP methods are respectively, and are calculated as follows:

[0174]

[0175]

[0176] in, and Calculated separately as follows:

[0177]

[0178]

[0179] in, This indicates the number of access points (APs) using the EP method for signal detection. This represents the average number of users whose signals were detected using the EP method.

[0180] For example, as Figure 1 The example shown is a decellularized massive MIMO system. Consider M = 36 access points (APs), with a uniform spacing of L. A =100m, each AP is configured with N=8 antennas, and K=20 users are randomly distributed within the AP coverage area. Each user requests service from an AP within a radius of R=100m centered on it. Channel h mk Large-scale fading coefficient β mk =tr(R) mk ) / N is:

[0181] β mk [dB] = -20.5 - 36.7 log 10 (d mk )+λ mk (27)

[0182] Where, d mk This represents the distance between the m-th access point (AP) and the k-th user. The shadow fading coefficient between the m-th AP and the k-th user is represented by the signal-to-noise ratio (SNR), which is defined as... Two modulation methods are used: Quadrature Phase Shift Keying (QPSK) and Quadrature Amplitude Modulation (16-QAM). The number of iterations of the iterative algorithm is set to T=6.

[0183] Based on the proposed scheme (MRC-EP), the system's bit error rate performance and computational complexity performance are obtained and compared with existing distributed MRC, EP, Linear Minimum Mean Square Error (LMMSE), and Large-MIMO Approximate Message Passing Algorithm (LAMA) methods. The performance of the centralized EP (Cen-EP) algorithm is used as a benchmark. The corresponding bit error rate comparison results are as follows: Figure 6 As shown, the comparison results of their computational complexity are as follows: Figure 7 As shown. (Through) Figure 6 It can be seen that the method proposed in this invention is superior to distributed MRC, LMMSE, and LAMA methods, and exhibits only a slight performance loss compared to the EP method. According to Figure 7 It can be seen that the hybrid distributed (MRC-EP) signal detection method proposed in this invention reduces the computational complexity by approximately 30% compared to the EP method. Combined with... Figure 6 and Figure 7 The hybrid distributed signal detection method proposed in this invention can achieve a trade-off between performance and computational complexity, verifying the effectiveness of the proposed method.

[0184] This invention employs hybrid distributed signal detection, effectively reducing the overhead of the fronthaul link between the AP and the CPU, as well as the computational load on the CPU, thus significantly reducing the amount of data transmitted through the fronthaul link. Considering the random distribution characteristics of users in decellularized massive MIMO systems, this invention adopts a user-centric dynamic cooperative cluster architecture. Each AP only provides services to users with good channel quality, effectively reducing interference between users within the group and the computational complexity of AP distributed signal detection. Furthermore, based on the actual overload ratio of the AP, this invention employs a flexible signal detection method, combining the complexity advantages of the MRC method and the performance advantages of the EP method. This ensures low computational complexity while maintaining distributed detection performance, meeting the implementation requirements of decellularized massive MIMO systems and demonstrating significant application value. Based on the hybrid information fusion rules, this invention designs a corresponding merging scheme at the CPU based on the characteristics of the distributed detection algorithm. For the EP method, it proposes an efficient nonlinear merging scheme, enhancing information fusion between APs and improving the performance of distributed detection. This invention includes different signal processing modules for the CPU and AP, which can be implemented through software, hardware, or a combination of both.

[0185] The above embodiments are only used to illustrate the design concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made based on the principles and design ideas disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A hybrid distributed signal detection method for a decellularized massive MIMO system, characterized in that, Includes the following steps: (1) Construct an uplink decellularized massive MIMO dynamic cooperative cluster communication model based on the user's transmitted signal and the channel estimation information obtained by the channel estimation module; (2) Calculate the overload ratio of each activated access point based on the number of antennas configured for each access point and the set of users served by that access point; (3) Multiple activated access points perform independent and parallel distributed signal detection: For each activated access point, an overload ratio threshold is preset for each activated access point. It is determined whether the calculated overload ratio is less than the preset overload threshold. If the calculated overload ratio is less than the preset overload threshold, the maximum ratio combining method is used to perform signal detection on the users served by the activated access point based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, so as to obtain the local signal detection results. Otherwise, the expected value propagation method is used to perform signal detection on the users served by the activated access point based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, so as to obtain the local signal detection results. (4) The signal detection module of each access point transmits the local detection results obtained in step (3) to the information merging and processing module of the centralized processing unit through the fronthaul link, and merges them using a hybrid linear-nonlinear information fusion rule to obtain the global signal detection results of each user. In step (4), the merging of signals using a hybrid linear-nonlinear information fusion rule to obtain the global signal detection results for each user specifically includes: For the Individual users, defined and The service number is respectively A set of access points for each user is analyzed using both the maximum ratio merging method and the expectation value propagation method. The local detection results of multiple access points obtained using the maximum ratio merging method are merged using a linear mixture method to obtain a first merged detection result. The local detection results of multiple access points obtained using the expectation value propagation method are then merged using a nonlinear mixture method to obtain a second merged detection result. Finally, the first and second merged detection results are merged using a linear mixture method to obtain a third merged detection result. Global signal detection results for each user; The method for obtaining the first merged detection result specifically includes: The first obtained using the maximum ratio merging method The user Local detection results of individual access points Represented as: （15） in, Indicates and Independent estimation error terms, with a mean of 0 and a variance of . The complex Gaussian distribution; No. The first merged detection result is obtained by linearly weighting the local detection results of multiple access points obtained by a user using the maximum ratio merging method. The expression is as follows: （16） （17） in, express The corresponding weighting factors satisfy ; The method for obtaining the second merged detection result specifically includes: When the preset maximum number of iterations is reached Then, based on the first value obtained using the expectation propagation method... The local detection results for each user will be accumulated to the average. and cumulative variance The data is transmitted to the centralized processing unit via the fronthaul link for nonlinear merging, resulting in: （18） （19） in, For the first Variance of individual users For the first The average of individual users; No. The posterior probability distribution of a user is represented as follows: （20） The posterior mean and posterior variance are expressed as follows: （21） in, Indicates the first The posterior mean of the nth user is the nth... The second merged detection result for each user Indicates the first Posterior variance of each user; The method for obtaining the user's global signal detection results specifically includes: A linear merging calculation is performed based on the first and second merged detection results to obtain the user's global signal detection result, expressed as follows: （22） in, and These represent the weighting factors for signal detection using the maximum ratio combining method and the expectation value propagation method, respectively, and their calculation formulas are as follows: （23） （24） in, and The calculation formulas are as follows: （25） （26） in, This represents the number of access points that use the expectation value propagation method for signal detection. This represents the average number of users who used the expected value propagation method to detect the signal.

2. The signal detection method according to claim 1, characterized in that, Step (1) specifically includes: No. The user and the first Service Identifier Matrix of Access Points Represented as: （3） in, express 3D identity matrix express Zero-dimensional matrix Represented as the first A set of access points that provide services to a user; By the A set of users served by an access point The expression is: （4） in, The size is , This represents the trace operation of a matrix; The expression for the communication model is: （5） in, Indicates the first The received signal of each access point , This represents the set of active access points. for 3D channel matrix, express Matrix transpose operation, for Each user sends a data signal, and each element of the signal is a constellation set. constellation points, It is an additive white Gaussian noise vector. Indicates that it follows the mean. Covariance is The complex Gaussian distribution.

3. The signal detection method according to claim 1, characterized in that, The formula for calculating the overload ratio is: （6） in, Indicates the first The overload ratio of each activated access point. For the first The number of antennas configured for each activated access point. For the first The size of the set of users served by each activated access point.

4. The signal detection method according to claim 1, characterized in that, In step (3), the maximum ratio combining method is used to perform signal detection on the users served by the activated access point based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, so as to obtain local signal detection results. Specifically, this includes: Local signal detection results Represented as: （7） in, Represents the Gram matrix, This represents the conjugate transpose operation of a matrix. This indicates retrieving the diagonal elements of the matrix. Represents the maximum ratio merge vector; Estimation error caused by using the maximum ratio combining method for signal detection Represented as: （8） in, express 3D identity matrix This represents the matrix inversion operation. Represented by matrix A diagonal matrix consisting of diagonal elements.

5. The signal detection method according to claim 1, characterized in that, In step (3), the expected value propagation method is used to perform signal detection on the users served by the activated access point based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, so as to obtain local signal detection results. Specifically, this includes: No. Joint distribution of signals transmitted and received by each access point Decomposed into: （9） in, Represents conditional distribution, express The prior distribution, Representing vectors The One element, Representing vectors The One element; make and They represent the first In the next iteration, from the variable node Passed to factor nodes The mean and variance, and These represent the factors from the nodes. Pass to variable node The mean and variance are calculated using the following formulas: （10） （11） in, and Representing the channel vectors respectively The and the One element, and Representing sets The Middle The posterior mean and posterior variance of each user are expressed as follows: （12） in, This represents the expectation operation. Indicates the first The posterior distribution probability of a user's constellation points is calculated using the following formula: （13） in, This represents the normalization operation. Indicates constellation points, Represents the prior probability of a constellation point, initialized to... , Indicates the size of the constellation set; All factor nodes To the variable node cumulative mean and cumulative variance They are represented as follows: （14） in, and They represent the first In the next iteration, from the factor node Pass to variable node Mean and variance, factor nodes Represents the likelihood function, variable node Represents the user.

6. A signal detection apparatus for implementing the signal detection method according to any one of claims 1-5, characterized in that, include: The signal receiving module is used to receive transmitted signals from the served users for the activated access point; wherein the transmitted signals include pilot signals and data signals; The radio frequency (RF) processing module is used to perform RF processing on the signals received by the signal receiving module; the RF processing includes filtering, low-noise amplification, and demodulation. The channel estimation module is used to estimate the channel information between each access point and the communication user based on the pilot signals sent by the user and using the minimum mean square error estimation method. The signal detection module, based on the data signal processed by the radio frequency processing module and the channel estimation information obtained by the channel estimation module, uses the maximum ratio combining method or the expectation value propagation method to perform signal detection on the users served by the access point, in order to obtain local signal detection results; and The information merging and processing module is used to merge the local signal detection results from each access point using a hybrid linear-nonlinear information fusion rule to calculate the global signal detection result for each user.

7. The signal detection device according to claim 6, characterized in that, The pilot signal received by the signal receiving module is represented as follows: （1） in, Indicates the first Each access point receives pilot signals. Indicates the first The access point and the first Channel vectors between users Indicates that it follows the mean. The covariance is The complex Gaussian distribution, Indicates assignment to the first Pilot signals for each user and , The transpose operation represents a vector or matrix. Indicates the pilot signal transmission time slot. express A Vaticonian white Gaussian noise matrix, wherein each element follows a mean of 0 and a variance of . The complex Gaussian distribution, This indicates the number of antennas at each access point; The channel estimation module uses the minimum mean square error estimation method to estimate the channel information between each access point and the communication user. The calculation formula is as follows: （2） in, express The estimated value, express 3D identity matrix Represents the conjugate operation of vectors or matrices.

Images