Signal Detection Method, Device and Electronic Device for mMIMO System

Abstract

The present invention provides a signal detection method, device and electronic device for an mMIMO system. The method includes: analyzing the received signal by using an FS-Net model to obtain an estimated value of the transmitted signal and a signal quantization solution; wherein, the received signal is obtained by transmitting the transmitted signal sent by the transmitting end through a channel to the receiving end; calculating the estimated value of the transmitted signal and the signal quantization solution based on the R3TS algorithm to obtain a final estimated solution of the transmitted signal. By inputting the received signal into the FS-Net model, the FS-Net model outputs an estimated value of the transmitted signal and a signal quantization solution with high precision. Based on the signal quantization solution as the search starting point of the R3TS algorithm, the final estimated solution of the transmitted signal is obtained. The R3TS algorithm further improves the accuracy of signal detection and has low complexity. By combining the FS-Net model and the R3TS algorithm for signal detection, low complexity and performance approximate to the optimal are achieved.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular, to a signal detection method, apparatus, and electronic device for a mMIMO system. Background Art

[0002] In order to meet the high-performance requirements of a massive multiple-input multiple-output (mMIMO) system, a receiver is usually equipped with multiple antennas and communicates with multiple transmitters simultaneously, which increases the complexity at the receiving end. Different data streams are transmitted from different antennas, causing interference between streams at the receiving end. Therefore, signal detection of multiple transmitted symbols in the receiving antennas becomes the key to realizing the MIMO prospect. The mMIMO considered in 5G wireless systems provides spatial degrees of freedom that enable it to support multiple users simultaneously, and the demand for high data rates leads to the use of higher modulation orders. The multiple combination of multiple users and higher data rates results in an exponential complexity of the search space, making it difficult for traditional signal detectors to be applied in practice. Traditional signal detectors usually process through algorithms based on mathematical models, and these traditional methods have various drawbacks, including high complexity, poor scalability, difficulty in online implementation, and inadaptability to dynamic environments.

[0003] Research shows that deep learning (DL) technology can significantly improve signal detection performance and is superior to classical signal detectors. Existing DL-based signal detectors generally expand a traditional iterative algorithm for signal detection into a network and train the parameters of the original algorithm. In a mMIMO system, there is still a large gap between the performance of such signal detectors and the performance of the theoretically optimal maximum likelihood (ML) detection algorithm; and when the complexity of the original algorithm is high or there are too many training parameters, problems such as complex neural networks and high offline training costs will occur, so the computational complexity is very high.

[0004] Therefore, existing DL-based signal detectors have problems of poor performance and high computational complexity in the practical application of full-rank transmission in a mMIMO system. Summary of the Invention

[0005] The present invention provides a signal detection method, apparatus, and electronic device for a mMIMO system, which are used to solve the defects of poor performance and high computational complexity in the prior art, and realize low-complexity signal detection with approximately optimal performance based on deep learning technology in a mMIMO system when the channel state information is known.

[0006] The present invention provides a signal detection method for a mMIMO system, including:

[0007] Analyze the received signal using the FS-Net model to obtain the estimated value of the transmitted signal and the signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitter being transmitted through the channel to the receiver for reception;

[0008] Calculate the estimated value of the transmitted signal and the signal quantization solution based on the R3TS algorithm to obtain the final estimated solution of the transmitted signal.

[0009] According to a signal detection method for an mMIMO system provided by the present invention, the step of calculating the estimated value of the transmitted signal and the signal quantization solution based on the R3TS algorithm to obtain the final estimated solution of the transmitted signal includes:

[0010] Based on the distance vector between the estimated value of the transmitted signal and the signal quantization solution, determine the positions of the symbols in the signal quantization solution that are misestimated;

[0011] Randomly correct the misestimated symbols to obtain the initial solution of the current RTS algorithm;

[0012] Run the RTS algorithm based on the initial solution to obtain the solution vector of the current RTS algorithm;

[0013] If the stop condition is not satisfied, return to execute the step of randomly correcting the misestimated symbols to obtain the initial solution of the current RTS algorithm; running the RTS algorithm based on the initial solution to obtain the solution vector of the current RTS algorithm;

[0014] If the stop condition is satisfied, determine the final estimated solution of the transmitted signal from all the current solution vectors.

[0015] According to a signal detection method for an mMIMO system provided by the present invention, the stop condition includes:

[0016] The cumulative number of times of currently running the RTS algorithm is greater than the preset number of times; or,

[0017] The cumulative number of times is less than the preset number of times and the result of the maximum likelihood cost function of the solution vector obtained by the current RTS algorithm is less than the preset threshold; or,

[0018] The cumulative number of times is less than the preset number of times, the result of the maximum likelihood cost function of the solution vector obtained by the current RTS algorithm is greater than or equal to the preset threshold, and the ratio of the number of different solution vectors currently obtained to the cumulative number of iterations of all the current RTS algorithms is less than or equal to the preset ratio.

[0019] A signal detection method for an mMIMO system provided by the present invention, the step of determining the symbol misestimated in the signal quantization solution based on the estimated value and the distance vector resolved from the signal quantity includes:

[0020] Obtain the distance vector of the estimated value and the signal quantization solution;

[0021] Determine the distance values in the distance vector that are greater than a preset error threshold;

[0022] The symbol corresponding to the distance value in the signal quantization solution is determined as the misestimated symbol.

[0023] A signal detection method for an mMIMO system provided by the present invention, the preset error threshold is:

[0024]

[0025] where γ represents the preset error threshold and SNR represents the signal-to-noise ratio.

[0026] A signal detection method for an mMIMO system provided by the present invention, the step of analyzing the received signal by using the FS-Net model to obtain the estimated value of the transmitted signal and the signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitting end being transmitted through the channel to the receiving end includes:

[0027] Divide the received signal into received pilot signals and received data signals;

[0028] Perform channel estimation on the received pilot signals to obtain a channel matrix;

[0029] Input the received data signals and the channel matrix into the FS-Net model to obtain the estimated value and the signal quantization solution.

[0030] A signal detection method for an mMIMO system provided by the present invention, the received pilot signals and the channel matrix are also used to form a pilot database with the pilot transmitted signals as the sample data set for updating the FS-Net model.

[0031] A signal detection method for an mMIMO system provided by the present invention, the method further includes:

[0032] Calculate the gap between the signal quantization solution and the final estimated solution;

[0033] If the gap is greater than a preset gap threshold, update the FS-Net model.

[0034] The present invention also provides a signal detection device for an mMIMO system, including:

[0035] The first acquisition module is configured to analyze the received signal by using the FS-Net model to obtain an estimated value of the transmitted signal and a signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitter being transmitted through the channel to the receiver for reception.

[0036] The second acquisition module is configured to calculate the estimated value of the transmitted signal and the signal quantization solution based on the R3TS algorithm to obtain the final estimated solution of the transmitted signal.

[0037] The present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the signal detection method of the mMIMO system as described in any one of the above is implemented.

[0038] The present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the signal detection method of the mMIMO system as described in any one of the above is implemented.

[0039] The present invention further provides a computer program product, including a computer program. When the computer program is executed by a processor, the signal detection method of the mMIMO system as described in any one of the above is implemented.

[0040] A signal detection method, device, and electronic device for an mMIMO system provided by the present invention, by inputting a received signal into the FS-Net model, the FS-Net model can output an estimated value of the transmitted signal and a signal quantization solution with high precision. Taking the output of the FS-Net model as the input of the R3TS algorithm, and based on the signal quantization solution as the search starting point of the R3TS algorithm to obtain the final estimated solution of the transmitted signal, the R3TS algorithm can further improve the precision of signal detection and has low complexity. Thus, by combining the FS-Net model and the R3TS algorithm for signal detection, low complexity and performance close to the optimal can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0042] Figure 1 It is a schematic flowchart of the signal detection method for the mMIMO system provided by the present invention;

[0043] Figure 2It is a schematic flowchart in the mMIMO system of the signal detection method for the mMIMO system provided by the present invention;

[0044] Figure 3 It is a schematic diagram of the network architecture of the FS-Net model provided by the present invention;

[0045] Figure 4 It is a schematic diagram of the operation process of the FS-Net model provided by the present invention;

[0046] Figure 5 It is a schematic diagram of the operation process of the R3TS algorithm provided by the present invention;

[0047] Figure 6 It is a schematic flowchart of the algorithm of the signal detection method for the mMIMO system provided by the present invention;

[0048] Figure 7 It is a schematic diagram of the structure of the signal detection device for the mMIMO system provided by the present invention;

[0049] Figure 8 It is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed implementation manners

[0050] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0051] It should be noted that in the description of the embodiments of the present invention, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitations, the element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0052] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of the same category, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" means at least one of the connected objects, and the character " / ", generally represents an "or" relationship between the associated objects before and after.

[0053] Before describing the signal detection method for the mMIMO system provided by the present invention, a brief introduction to the mMIMO system to which this method is applied is given first. Consider the uplink of a multi-user MIMO system, where the base station is equipped with N r receive antennas, one user is equipped with one transmit antenna, and there are N t users in total. The received signal vector can be given by the expression of the complex signal model:

[0054]

[0055] where the transmit symbol vector i = 1, 2, …, N t , where P i is the average power of the transmit symbol. The transmit symbol is independently and randomly generated by a complex constellation A with points. Let the set of all possible transmission vectors be which is an N -dimensional complex set composed of t types of vectors, that is In the above formula, is a vector of independent and identically distributed additive white Gaussian noise (AWGN, Additive White Gaussian Noise) samples, the channel matrix H is a matrix of size N ×N r ×N t formed by where

[0056] represents the complex channel gain between the jth transmit antenna and the ith receive antenna. The channel matrix H is sampled from an i.i.d complex Gaussian normal distribution or generated by simulating a typical 5G communication channel.

[0057]

[0058] Among them, ΔH represents the error matrix, which is composed of to form a matrix of size N r ×N t .

[0059] Let s, y, n, and H represent the M×1-dimensional equivalent real transmitted signal, the N×1-dimensional equivalent real received signal, the N×1-dimensional equivalent real AWGN noise signal, and the N×M-dimensional equivalent real channel matrix, respectively, and M = 2N t and N = 2N r , where:

[0060]

[0061] Here and represent the real part and the imaginary part of a complex vector or matrix, respectively. Therefore, the complex signal model can be converted into an equivalent real signal model:

[0062] y = Hs + n

[0063] Let the set composed of all possible transmission vectors be A M , which is a set of M-dimensional complex numbers composed of Q M vectors, that is, s ∈ A M , where A is a real-valued symbol set. Therefore, it is set that there are |A| M elements in the set A M . It can be understood that an equivalent real signal model is adopted to replace the complex signal model for the convenience of calculation.

[0064] It should be understood that the goal of signal detection is to estimate the transmitted signal vector s from the received signal vector y. During the data transmission process of the mMIMO system, signal detection generally occurs before the received signal is demodulated.

[0065] Next, the signal detection method of the mMIMO system of the present invention will be described in conjunction with Figures 1-6 .

[0066] As Figure 1 shown, the signal detection method of the mMIMO system provided by the present invention includes:

[0067] Step 110: Analyze the received signal by using the FS-Net model to obtain the estimated value of the transmitted signal and the signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitting end being transmitted through the channel to the receiving end for reception.

[0068] Specifically, the FS-Net (the Fast-convergence Sparsely Connected Detection Network) model can be a trained FS-Net model. When the received signal is input into the FS-Net model, it can generate a highly accurate and reliable estimated value of the transmitted signal and a signal sparsity solution.

[0069] Step 120: Calculate the estimated value of the transmitted signal and the signal sparsity solution based on the R3TS algorithm to obtain the final estimated solution of the transmitted signal.

[0070] Specifically, the overall idea of the R3TS (Random-restart Reactive Tabu Search) algorithm is to randomly generate different initial solutions, run multiple RTS (Reactive Tabu Search) algorithms, and then select the best solution vector from the solution vectors obtained by running multiple RTS algorithms as the estimated solution of the transmitted signal.

[0071] A signal detection method for an mMIMO system provided by the present invention inputs the received signal into the FS-Net model. The FS-Net model can output a highly accurate estimated value of the transmitted signal and a signal sparsity solution. The output of the FS-Net model is used as the input of the R3TS algorithm, and the signal sparsity solution is used as the search starting point of the R3TS algorithm to obtain the final estimated solution of the transmitted signal. The R3TS algorithm can further improve the accuracy of signal detection and has a low complexity. Therefore, by combining the FS-Net model and the R3TS algorithm for signal detection, low complexity and approximately optimal performance can be achieved.

[0072] As Figure 2 shown, a signal detection method for an mMIMO system provided by the present invention is applied before receiving signal demodulation.

[0073] The following will describe in detail a signal detection method for an mMIMO system provided by the present invention from two aspects: the FS-Net model and the R3TS algorithm.

[0074] 1. FS-Net model:

[0075] In the field of signal detection, the ML (Maximum Likelihood) detection algorithm is recognized as the optimal detection algorithm. It globally searches the received signal over all possible transmitted signal domains and finds the transmitted signal with the minimum distance from the received signal as the original transmitted signal. Its estimation formula is:

[0076]

[0077] The gradient descent method can execute the ML detection algorithm in multiple iterations, that is, optimize the valuation formula of ML, and the objective function is ‖y - Hs‖ 2 The gradient with respect to s is:

[0078]

[0079] Update s through the gradient descent method:

[0080]

[0081] The learning rate δ in the formula determines the magnitude of the moving speed towards the optimal result, and l represents the l-th iteration.

[0082] This idea is equivalent to using the gradient descent method with a step size of δ to optimize the estimated value of the transmitted signal s by minimizing ‖y - Hs‖ in multiple iterations 2 and then project the predicted signal onto the constellation set A M That is, find the point closest to it in the constellation set A of the transmitted signal M to obtain the estimated value of s

[0083] Since the computational complexity of the ML detection algorithm increases exponentially, it is difficult to apply it to practical systems.

[0084] The FS-Net model of the present invention is based on solutions such as mimicking projected gradient descent. This idea will lead to iterations in the following form:

[0085]

[0086] where is the estimated value of the l-th iteration, ∏(·) is the projection operator, and δ [l-1] is the step size. Each iteration is H T y, a linear combination of, and then through a non-linear function. In a deep learning network, there is the following analogy: ∏(·) is equivalent to the non-linear activation function of each layer, and are the input and output of the l-th layer network respectively. The network architecture of the FS-Net model is as Figure 3 shown.

[0087] In the FS-Net model, set the non-linear activation function of each layer to:

[0088]

[0089] Among them, when QPSK modulation is adopted, q = 1, Ω = {0} to ensure that the symbols in are within the range of [-1, 1]; when 16-QAM modulation is adopted, q = 3, Ω = {-2, 0, 2} to ensure that the symbols in are within the range of [-3, 3]. The final predicted signal

[0090] The FS-Net model takes into account the correlation between and s in the loss function to accelerate learning convergence. The loss function is:

[0091]

[0092] where Based on the inequality When the equal sign holds.

[0093] It should be understood that the FS-Net model is a trained FS-Net model, that is, a model constructed based on FS-Net, and has a certain generalization ability, and can output the estimated value of the transmitted signal and the signal quantization solution according to the received signal input.

[0094] In one embodiment, the training steps of the FS-Net model include: initializing the network parameters of FS-Net; training based on the input training data set and defined hyperparameters, and updating the network parameters to obtain the FS-Net model.

[0095] Specifically, input the offline training data set {s, y, H}, and define the hyperparameters learning rate α F and the number of training iterations n F . Randomly initialize the network parameters θ of FS-Net, let i = 1, and use the training data set to train FS-Net. The specific process is:

[0096]

[0097] l = 1 → L

[0098]

[0099]

[0100]

[0101] Update the network parameters of FS-Net:

[0102]

[0103] Among them, i represents the current iteration number. It can be understood that each time the network parameters are updated during training, i = i + 1, and the training continues to iteratively update the network parameters. When i > n F , the training ends. Finally, the trained FS-Net model is output based on the updated network parameters, that is, the FS-Net model.

[0104] As Figure 4 shown, in one embodiment, the steps of inputting the received signal into the FS-Net model to obtain the estimated transmitted signal value and the signal quantization solution include:

[0105] Step 410, dividing the received signal into a received pilot signal and a received data signal.

[0106] Step 420, performing channel estimation on the received pilot signal to obtain a channel matrix.

[0107] Step 430, inputting the received data signal and the channel matrix into the FS-Net model to obtain the estimated value and the signal quantization solution.

[0108] In this embodiment, the received signal y can be divided into a received pilot signal y p and a received data signal y d . The received pilot signal y p can be used to calculate the channel matrix H through channel estimation; inputting the received data signal y d and the channel matrix H into the FS-Net model can obtain a relatively accurate initial estimated transmitted signal value and the corresponding signal quantization solution, that is, and Put and into the optimized R3TS algorithm to obtain the final estimated solution of the transmitted signal for signal detection

[0109] In one embodiment, the received pilot signal and the channel matrix are also used to form a pilot database with the pilot transmitted signal as the sample data set for updating the FS-Net model.

[0110] Specifically, the received pilot signal t p , the channel matrix H, and the known pilot transmitted signal s p can form a pilot database {s p , y p , H} as the sample data set for updating the FS-Net model.

[0111] In one embodiment, the method further includes: calculating the gap between the semaphore solution and the final estimated solution of the transmitted signal; if the gap is greater than a preset gap threshold, updating the FS-Net model.

[0112] Specifically, calculate the semaphore solution output by the FS-Net model and the final estimated solution of the transmitted signal output by the subsequent R3TS algorithm of the gap If the gap is greater than the preset gap threshold (i.e., ), then update the FS-Net model.

[0113] It should be understood that this embodiment utilizes the characteristic that the pilot information is known, and in practical applications, it can continuously collect the sample data set {s p , y p , H}, and uses the collected sample data set to update the parameters of the FS-Net model, so that the FS-Net model can more accurately recover the transmitted signal and improve the generalization ability of the FS-Net model.

[0114] In one embodiment, the method further includes: updating the FS-Net model at every preset time interval. That is to say, the FS-Net model can be updated regularly to improve the generalization ability of the FS-Net model.

[0115] It can be understood that the above two update mechanisms of the FS-Net model can be used alone or in combination.

[0116] In one embodiment, the update step of the FS-Net model includes: loading the network parameters of the current FS-Net model; training based on the input sample data set and defined hyperparameters, and updating the network parameters to obtain an updated FS-Net model.

[0117] Specifically, input the sample data set {s p , y p , H}, define the hyperparameters learning rate α′ F , training iteration times n′ F , randomly initialize the network parameters θ′ of the FS-Net, let i = 1, and use the training data set to train the FS-Net and update the network parameters of the FS-Net:

[0118]

[0119] After n′ FAfter the training iteration, the training ends. Finally, the trained FS-Net model is output based on the updated network parameters, that is, the updated FS-Net model. Thereafter, the updated FS-Net model is used for signal detection.

[0120] 2. R3TS Algorithm

[0121] The R3TS algorithm provided by the present invention has been optimized in two aspects:

[0122] (1) Optimize the random initial solution of R3TS: By randomly correcting the signal quantity, the unreliable detection symbols in the solution are resolved, and Based on this, multiple R3TS random initial solutions closer to the final estimated solution are generated. These high-precision random initial solutions can greatly reduce the search times of R3TS, thereby greatly reducing the complexity of the overall signal detection method on the basis of approaching the theoretical optimal R3TS performance.

[0123] (2) Optimize the number of iterations of R3TS: The R3TS algorithm of the present invention considers the signal-to-noise ratio knowledge and proposes a stopping criterion (stopping condition) for the algorithm. The R3TS algorithm proposes an effective adaptive stopping criterion instead of simply stopping the algorithm by the number of iterations. When the initial solution may be accurate, the number of iterative searches is reduced, thus reducing the complexity of the R3TS algorithm.

[0124] Such as Figure 5 As shown, in one embodiment, the step of calculating the final estimated solution of the transmitted signal based on the R3TS algorithm for the estimated value and the signal quantity solution includes:

[0125] Step 510, determine the symbols misestimated in the signal quantity solution based on the distance vector between the estimated value and the signal quantity solution.

[0126] Specifically, according to the estimated value of the transmitted signal output by the FS-Net model And the signal quantity solution Calculate the distance vector between them The elements in e represent And The distance of specific symbols between them, predict the accuracy of the quantity solution , by checking the elements of e, to determine whether there is a high error probability in the symbols of , and regard the symbols with high error probability in As the predicted error symbols.

[0127] In one embodiment, the step of determining the symbols in the signal quantization solution that are misestimated based on the estimated value of the transmitted signal and the distance vector resolved by the signal quantity includes: obtaining the estimated value of the transmitted signal and the distance vector resolved by the signal quantity; determining the distance values in the distance vector that are greater than a preset error threshold; and determining the symbols corresponding to the distance values in the signal quantization solution as the misestimated symbols.

[0128] Specifically, if e m exceeds the predefined error threshold γ, then the m-th symbol in is determined as a predicted error symbol. The preset error threshold is inversely proportional to the signal-to-noise ratio. Since 0 ≤ e m ≤ 1, the preset error threshold γ is:

[0129]

[0130] where γ represents the preset error threshold and SNR represents the signal-to-noise ratio.

[0131] It should be understood that is the output result of the FS-Net model, is the hard decision quantization result, is closer to the original transmitted signal vector than . The overall accuracy of the signal quantization solution output by the FS-Net model is relatively high, but there will also be some symbols in that are misestimated. Among these symbols, the ones most likely to be misestimated are the symbols corresponding to the elements with larger e values.

[0132] For example, in a 2×2 MIMO system under QPSK modulation, if the estimated value of the transmitted signal is then the corresponding signal quantization solution is then e = [0.95, 0.3, 0.1, 0.75] T , and the corresponding corrected quantization solution is From this, it can be obtained that S1 and S4 are misestimated. Therefore, it can be considered that the symbols exceeding the predefined error threshold γ are misestimated symbols.

[0133] Find the position P of the elements in e that are greater than γ. The position P is recognized as the position of the misestimated symbol.

[0134] Step 520, randomly correct the misestimated symbols to obtain the initial solution of the current RTS algorithm.

[0135] Specifically, let m e represent the number of predicted error symbols in, and let S be the correction The set of all candidates obtained. Where S (k) , 1 < k < m e is the subset obtained by correcting k symbols of.

[0136] It should be understood that the present invention does not concern how many vectors there are in the set S, but obtains random different initial solutions of R3TS from random S, and runs the RTS algorithm based on the random different initial solutions respectively. Let S (j) , j = 1, 2, 3,... represents the random initial vector for running the j-th RTS algorithm, where and

[0137] The specific operation to obtain the initial solution S of each RTS (j) is as follows:

[0138] Let the current The positions of the elements in e that exceed the preset error threshold γ are P = [P1, P2,..., P p T . Randomly select some values of P to obtain P (j) = [P1, P2,..., P p(j) T , and find the symbols at the positions corresponding to these values of P in (j) and correct them to obtain S (j) .

[0139] Specifically, the correction operation is to quantize the symbol to be corrected to the second nearest constellation value. For example, in QPSK modulation, the symbol values in are {±1}, so the correction is to invert the element values at the corresponding positions; in 16-QAM modulation, for another example, the symbol values in are {±1, ±3}, and the correction operation is more troublesome. Assume the symbol at position m in needs to be corrected, then the corrected symbol is:

[0140]

[0141] The solution generated after correcting the symbol is closer to the original transmitted signal vector, and the search path becomes shorter based on the initial solution obtained after correction, so the algorithm complexity is reduced.

[0142] Step 530, run the RTS algorithm based on the initial solution to obtain the solution vector of this RTS algorithm.

[0143] In one embodiment, the adaptive cutoff factor can be calculated according to the position P ​​

[0144] It should be understood that in one RTS algorithm, iterative search is also performed, and the best solution vector of the current RTS algorithm is obtained from the candidate solution vectors obtained through multiple iterative searches. To reduce the number of redundant search iterations, an adaptive ET (Early-termination) criterion is proposed. Let m e represent the number of predicted error symbols in e . If m = 0, then there are no predicted error symbols in , which means that the probability that e is already the optimal solution is very high. In this case, no further search iteration is required, that is, the number of search iterations I = 0. If m is very small, then there is a high probability that only a few symbols in e are misestimated, and only a small number of corrections are needed. Therefore, in this embodiment, only a very small I is required. When m is much greater than 1, sufficient search iterations are required to ensure optimal performance. Therefore, an adaptive cutoff factor is proposed, which depends on

[0145]

[0146] where M is the number of symbols in , and m e is the number of elements of P (the number of predicted error symbols in ), and μ is used to ensure that there are sufficient search iterations.

[0147] Step 540, determine whether the stop condition is satisfied after running the current RTS algorithm; if the stop condition is not satisfied, return to execute step 520; if the stop condition is satisfied, execute step 550.

[0148] In one embodiment, the stop condition may include the following three types:

[0149] The cumulative number of times the current RTS algorithm is run is greater than the preset number; or,

[0150] The cumulative number of times is less than the preset number and the result of the maximum likelihood cost function of the solution vector obtained by the current RTS algorithm is less than the preset threshold; or,

[0151] The cumulative number of times is less than the preset number of times, the result of the maximum likelihood cost function of the solution vector obtained by the current RTS algorithm is greater than or equal to the preset threshold, and the ratio of the number of different solution vectors currently obtained to the cumulative number of iterations of all the current RTS algorithms is less than or equal to the preset ratio.

[0152] Specifically, by setting three stopping conditions and defining three parameters MAX, Θ, p, the number of iterative searches is reduced and the complexity of the R3TS algorithm is decreased.

[0153] The first stopping condition is that the cumulative number of times of running the RTS algorithm reaches the preset number of times (the maximum number of runs MAX). If the maximum number of runs is reached, the search is stopped and the final estimated solution of the transmitted signal is determined from all the current solution vectors.

[0154] The second stopping condition is that when the cumulative number of times of running the RTS algorithm does not reach the maximum number of runs, it is judged whether the result of the ML cost function of the solution vector is less than the preset threshold. If it is less than the preset threshold, the search is stopped, and the solution vector with the current result of the ML cost function less than the preset threshold is the final estimated solution of the transmitted signal.

[0155] Specifically, the idea of the R3TS algorithm is to have different random initial solutions, run multiple RTS algorithms, and select the best solution vector from multiple solution vectors. It takes into account the knowledge of signal-to-noise ratio and proposes the stopping conditions of the algorithm. The ML criterion is to minimize If the solution vector is the same as the transmission vector s, the ML cost function is ‖n‖ 2 (Since n is a vector of independent and identically distributed additive white Gaussian noise (AWGN) samples, where So ‖n‖ 2 is a non-central chi-square distribution with a mean of and a variance of ), and empirically set This algorithm determines that when the ML cost function is less than Θ, the estimated value of the transmitted signal is accurate enough. Therefore, this solution vector can be used as the final estimated solution of the transmitted signal.

[0156] The third stopping condition is that when the cumulative number of times of running the RTS algorithm does not reach the maximum number of runs and the result of the ML cost function of the solution vector is greater than or equal to the preset threshold, it is judged whether the ratio of the number of different solution vectors currently obtained to the cumulative number of iterations of all the current RTS algorithms is less than or equal to the preset ratio p.

[0157] Specifically, K represents the number of iterations that have been performed so far (the accumulation of the number of iterations of each RTS algorithm), and L represents the number of different solution vectors so far (the accumulation of the solution vectors obtained by each RTS). If L / K ≤ p, the search is stopped, and the final estimated solution of the transmitted signal is determined from all the current solution vectors. The motivation for the third stopping condition is to reduce the implementation complexity in the case where ‖n‖ 2 is just greater than Θ.

[0158] Step 550, determine the final estimated solution from all the current solution vectors.

[0159] Combined Figure 6 , the detailed algorithm flow of a signal detection method for an mMIMO system provided by the present invention is as follows:

[0160] Input: {y d , H}, I UB , ε, max_rep, β, MAX, Θ, p; Output:

[0161] (1) Input the received signal into the FS-Net model. Obtain and

[0162] (2)

[0163] (3)

[0164] (4) Find the positions P of the elements in e that are greater than γ

[0165] (5) m e is the number of elements of P

[0166] (6) Initialize Put c into the taboo list L (the taboo list L stores the different solution vectors so far)

[0167] (7)

[0168] (8) Calculate the adaptive cut-off factor Therefore Initialize the taboo period P = I UB

[0169] (9) K = L = 0, where K represents the total number of iterations that have been performed so far (the accumulation of the number of iterations of multiple RTS algorithms), and L represents the number of different solution vectors so far (the accumulation of the solution vectors obtained by multiple RTS algorithms, that is, the number of solution vectors in the taboo list)

[0170] (10) number = 1 (the number of times the RTS algorithm runs)

[0171] (11) Random correction the symbol at position P to obtain the initial solution S of each RTS algorithm (j) (S (j) where j represents the jth RTS algorithm being performed, and

[0172] (12) i = 1, count = 0, q = 0, l p = 0 (i records the number of iterations, count records the number of consecutive iterations without finding a better solution, q records the number of times the new solution is a duplicate solution, and l p records the number of iterations elapsed since the last change in the P value)

[0173] (13) if i ≤ P and count ≤ I e then

[0174] (14) Find the adjacent nodes of c in N(c), where N(c) represents all non-taboo neighborhoods of the current c, N(c) = {z ∈ A M \L, |z - c| = θ min}, A M \L represents the constellation set of the taboo vectors retained in the taboo list L, and θ M is the minimum distance between two constellation points on the plane min

[0175] (15) (Check all non-taboo neighborhoods N(c) of the current candidate c to find the best neighbor with the minimum ML metric φ(s) )

[0176] (16) Update the current candidate solution

[0177] (17) (c represents the new candidate solution obtained from the search in this iteration, represents the best candidate solution before the search in this iteration)

[0178] (18) Update the best candidate

[0179] (19) count = 0

[0180] (20) else

[0181] (21) count = count + 1

[0182] (22) end if

[0183] (23) Put c into the tabu list L

[0184] (24) i = i + 1

[0185] (25) If i > P or count > I e , jump to step 36; otherwise go to the next step

[0186] (26) Record the iteration number c.No and the ML cost function value c.ML corresponding to the current c

[0187] (27) if there exists a solution vector in the tabu list L that has a duplicate solution relationship with the current c then

[0188] (28) Calculate the interval length l at which the current duplicate solution appears c , that is, the difference between the two iteration numbers; update the average length (if l rep ≠0 then l rep =(l c +l rep ) / 2 else l rep =l c )

[0189] (29) Increase the tabu period P = P + 1, record the number of times q that the duplicate solution appears = q + 1

[0190] (30) If q > max_rep, jump to step 36; otherwise go to the next step

[0191] (31) l p =0

[0192] (32) else

[0193] (33) l p =l p +1, if l p >βl rep then P = max{1, P, -1}

[0194] (34) end if

[0195] (35) Jump to step 13

[0196] (36) The optimal solution found from the current tabu list L where satisfies

[0197] (37) Calculate L (the number of solution vectors in the tabu list) and K (the cumulative iteration number K = K + i) after each RTS algorithm ends

[0198] (38) If number > MAX, then jump to step 42

[0199] (39) Then this solution vector is used as the final solution, and jump to step 43

[0200] (40) Else if L / K ≤ p, then jump to step 42

[0201] (41) Else number = number + 1, and return to step 11

[0202] (42) Output the optimal solution found so far (send signal final estimate solution)

[0203] (43) end

[0204] In the above algorithm flow, steps 1 - 12 are to obtain the initial solution and initialize the parameters; steps 13 - 37 are the search phase of the RTS algorithm each time; steps 38 - 43 are the stop criteria (stop conditions) of the R3TS algorithm.

[0205] The signal detection method of the mMIMO system provided by the present invention combines the deep learning network FS - Net and the traditional R3TS signal detection algorithm, reducing the computational complexity while improving the performance. Compared with the existing signal detection technologies, the present invention has the following improvements:

[0206] (1) First, the R3TS algorithm randomly generates different initial solutions, runs multiple RTS algorithms, and then selects the best solution vector from multiple solution vectors, so it can improve the algorithm performance. The present invention generates "multiple random initial solutions" with higher accuracy in R3TS. The present invention is not a simple combination of the FS - Net model and the R3TS algorithm. In the traditional R3TS algorithm, different initial solutions are generated randomly. Although different initial solutions of R3TS can be generated by making random changes based on the initial solutions generated by FS - Net in the first stage, this method cannot guarantee the accuracy of these initial solutions and has no theoretical basis. Therefore, the present invention proposes that first, predict the reliable or unreliable detection symbols in the initial solution according to the output of FS - Net, and obtain different initial solutions of R3TS by randomly correcting the unreliable symbols of this initial solution. The initial solutions obtained in this way are, theoretically speaking, more accurate than the output of FS - Net. Therefore, the algorithm complexity can be further reduced while improving the algorithm performance.

[0207] (2) Second, the R3TS algorithm proposes a more effective stop criterion. Using the knowledge of the ML cost function, when ‖y - Hs‖ 2When it is less than a certain value, it is determined that the current solution is accurate enough to stop the algorithm early, avoiding unnecessary searches. Therefore, it can improve the algorithm performance while reducing the algorithm complexity.

[0208] (3) The R3TS algorithm also proposes richer and more comprehensive algorithm stopping criteria, such as restricting the number of repeated solutions, restricting the length of the taboo list, balancing the current number of iterations and the current number of solutions, etc. The purpose is to stop the algorithm in time to avoid inefficient searches when it is likely that no better solution can be obtained in future searches. Therefore, the algorithm complexity can be reduced.

[0209] The signal detection device of the mMIMO system provided by the present invention will be described below. The signal detection device of the mMIMO system described below can be mutually corresponding and referred to the signal detection method of the mMIMO system described above.

[0210] As Figure 7 shown, the present invention provides a signal detection device for an mMIMO system, including:

[0211] A first acquisition module 710, configured to analyze the received signal by using the FS-Net model to obtain an estimated value of the transmitted signal and a signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitting end being transmitted through a channel to the receiving end for reception;

[0212] A second acquisition module 720, configured to calculate the estimated value of the transmitted signal and the signal quantization solution based on the R3TS algorithm to obtain a final estimated solution of the transmitted signal.

[0213] Figure 8 An entity structure diagram of an electronic device is exemplified. As Figure 8 shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communication interface 820, and the memory 830 complete mutual communication through the communication bus 840. The processor 810 can call the logical instructions in the memory 830 to execute the signal detection method of the mMIMO system, and the method includes: analyzing the received signal by using the FS-Net model to obtain an estimated value of the transmitted signal and a signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitting end being transmitted through a channel to the receiving end for reception; calculating the estimated value and the signal quantization solution based on the R3TS algorithm to obtain a final estimated solution of the transmitted signal.

[0214] In addition, when the logical instructions in the above-mentioned memory 830 can be implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0215] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the signal detection method of the mMIMO system provided by the above-mentioned various methods. The method includes: analyzing the received signal by using the FS-Net model to obtain an estimated value of the transmitted signal and a signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitting end being transmitted through a channel to the receiving end for reception; calculating the estimated value and the signal quantization solution based on the R3TS algorithm to obtain a final estimated solution of the transmitted signal.

[0216] In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the signal detection method of the mMIMO system provided by the above-mentioned various methods. The method includes: analyzing the received signal by using the FS-Net model to obtain an estimated value of the transmitted signal and a signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitting end being transmitted through a channel to the receiving end for reception; calculating the estimated value and the signal quantization solution based on the R3TS algorithm to obtain a final estimated solution of the transmitted signal.

[0217] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0218] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0219] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A signal detection method for an mMIMO system, characterized in that, Comprising: Analyze the received signal using the FS-Net model to obtain an estimated value of the transmitted signal and a signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitter being transmitted through the channel to the receiver for reception; Calculate the estimated value and the signal quantization solution based on the R3TS algorithm to obtain the final estimated solution of the transmitted signal; The step of calculating the estimated value and the signal quantization solution based on the R3TS algorithm to obtain the final estimated solution of the transmitted signal includes: Determine the symbols in the signal quantization solution that are misestimated based on the distance vector between the estimated value and the signal quantization solution; Randomly correct the misestimated symbols to obtain the initial solution of the current RTS algorithm; Run the RTS algorithm based on the initial solution to obtain the solution vector of the current RTS algorithm; Judge whether the stop condition is satisfied after running the current RTS algorithm; If the stop condition is not satisfied, return to execute the step of randomly correcting the misestimated symbols to obtain the initial solution of the current RTS algorithm; running the RTS algorithm based on the initial solution to obtain the solution vector of the current RTS algorithm; If the stop condition is satisfied, determine the final estimated solution from all the current solution vectors; The step of determining the symbols in the signal quantization solution that are misestimated based on the distance vector between the estimated value and the signal quantization solution includes: Obtain the distance vector between the estimated value and the signal quantization solution; Determine the distance values in the distance vector that are greater than the preset error threshold; The symbols corresponding to the distance values in the signal quantization solution are determined as the misestimated symbols; The step of analyzing the received signal using the FS-Net model to obtain an estimated value of the transmitted signal and a signal quantization solution includes: Divide the received signal into a received pilot signal and a received data signal; Perform channel estimation on the received pilot signal to obtain a channel matrix; Input the received data signal and the channel matrix into the FS-Net model to obtain the estimated value and the signal quantization solution.

2. The signal detection method for an mMIMO system according to claim 1, characterized in that, The stop condition includes: The cumulative number of times of currently running the RTS algorithm is greater than the preset number of times; or, The cumulative number of times is less than the preset number of times and the result of the maximum likelihood cost function of the solution vector obtained by the current RTS algorithm is less than the preset threshold; or, The cumulative number of times is less than the preset number of times, the result of the maximum likelihood cost function of the solution vector obtained by the current RTS algorithm is greater than or equal to the preset threshold, and the ratio of the number of different solution vectors currently obtained to the cumulative number of iterations of all the current RTS algorithms is less than or equal to the preset ratio.

3. The signal detection method for an mMIMO system according to claim 1, characterized in that, The preset error threshold is: ； Among them, represents the preset error threshold, represents the signal-to-noise ratio.

4. The signal detection method for an mMIMO system according to claim 1, characterized in that, The received pilot signal and the channel matrix are also used to form a pilot database with the pilot transmitted signal and serve as a sample data set for updating the FS-Net model.

5. The signal detection method for an mMIMO system according to claim 1, characterized in that, The method further includes: Calculate the gap between the signal quantization solution and the final estimated solution; If the gap is greater than the preset gap threshold, update the FS-Net model.

6. A signal detection device for an mMIMO system, characterized in that, Comprising: The first acquisition module is used to analyze the received signal by using the FS-Net model to obtain an estimated value of the transmitted signal and a signal quantization solution; wherein, the received signal is obtained by the transmitted signal sent by the transmitting end being transmitted through the channel to the receiving end for reception; The second acquisition module is used to calculate the estimated value and the signal quantization solution based on the R3TS algorithm to obtain the final estimated solution of the transmitted signal; The calculating the estimated value and the signal quantization solution based on the R3TS algorithm to obtain the final estimated solution of the transmitted signal includes: Determining the symbols in the signal quantization solution that are misestimated based on the distance vector between the estimated value and the signal quantization solution; Randomly correcting the misestimated symbols to obtain the initial solution of the current RTS algorithm; Running the RTS algorithm based on the initial solution to obtain the solution vector of the current RTS algorithm; Judging whether the stop condition is satisfied after running the current RTS algorithm; If the stop condition is not satisfied, return to execute the step of randomly correcting the misestimated symbols to obtain the initial solution of the current RTS algorithm; running the RTS algorithm based on the initial solution to obtain the solution vector of the current RTS algorithm; If the stop condition is satisfied, determine the final estimated solution from all the current solution vectors; The determining the symbols in the signal quantization solution that are misestimated based on the distance vector between the estimated value and the signal quantization solution includes: Obtaining the distance vector between the estimated value and the signal quantization solution; Determining the distance values in the distance vector that are greater than the preset error threshold; The symbols corresponding to the distance values in the signal quantization solution are determined as the misestimated symbols; The analyzing the received signal by using the FS-Net model to obtain the estimated value of the transmitted signal and the signal quantization solution includes: Dividing the received signal into a received pilot signal and a received data signal; Performing channel estimation on the received pilot signal to obtain a channel matrix; Inputting the received data signal and the channel matrix into the FS-Net model to obtain the estimated value and the signal quantization solution.

7. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, When the processor executes the program, it implements the signal detection method of the mMIMO system according to any one of claims 1 to 5.

Images