Active user detection and channel estimation method and device

Abstract

The present application discloses a method and apparatus for detecting active users and estimating channels. The method comprises: determining a channel model and a received signal model in a human coexistence network based on signals received by a base station; utilizing the spatial correlation between the channels based on the channel model and the received signal model to derive the posterior probability density of the channel in the human coexistence network and the total likelihood function of the received signal; utilizing the type-II maximum likelihood estimation method based on the posterior probability density and the total likelihood function to obtain a cost function; and detecting active users in the human coexistence network and estimating the channel based on the cost function. The present application solves the technical problem of poor communication performance caused by the inability to utilize the spatial correlation between different antennas.

Description

Technical Field

[0001] The present application relates to the field of communications, and in particular to a method and apparatus for active user detection and channel estimation. Background Art

[0002] Massive Machine Type of Communication (mMTC) in human-machine coexistence networks plays a vital role in 5G / B5G mobile wireless communication networks. It can provide massive connections and ultra-low latency for large-scale Internet of Things (IoT) devices. mMTC mainly focuses on the uplink sporadic small packet communication of large-scale IoT devices. Sporadic small packet communication means that only a small number of devices are active at the same time, and the active users send short packet data with low transmission rate to the base station. However, in actual use, the complex handshake link in the scheduling process will cause excessive signaling overhead, and the allocation of limited orthogonal resources to a large number of machine-type communication devices will result in insufficient time or frequency resources. Therefore, mMTC still has room for improvement.

[0003] In networks where humans and machines coexist, non-orthogonal multiple access (NOMA) and unlicensed access have become research hotspots to support large-scale IoT device connections. NOMA allows multiple users to share the same resource block. Therefore, this technology can support massive connections even with limited resources. As mentioned above, applying NOMA technology to mMTC scenarios has alleviated the resource shortage issue to some extent. However, the signaling overhead generated by the complex scheduling process is enormous compared to the sporadic transmission of short data packets. Unlicensed access can be used to address this excessive signaling overhead. Unlicensed access allows users to skip the complex handshake process during scheduling and transmit information directly to the base station. Therefore, unlicensed NOMA systems are well-suited for mMTC scenarios. Furthermore, multiple-input multiple-output (MIMO), a long-standing research technology, has become highly mature. MIMO not only significantly improves system throughput, but also extends coverage and reduces power consumption. In recent years, MIMO and NOMA have been frequently combined to further improve spectrum efficiency. In the unlicensed MIMO-NOMA system, mMTC devices send information directly to the base station without scheduling, so it is necessary to find a small number of active users among a large number of potential users and estimate the transmission channel.

[0004] Currently, there has been some research on active user detection and channel estimation, with compressed sensing-based detection methods being particularly common. Compressed sensing exploits signal sparsity to accurately reconstruct the original signal from a limited number of random mappings. Therefore, compressed sensing-based methods are well-suited for mMTC scenarios characterized by sporadic transmission. With the advancement of machine learning, compressed sensing-based machine learning methods have provided new approaches to solving active user detection and channel estimation problems. Sparse Bayesian learning is one of the most innovative research areas. Based on a Bayesian hierarchical model, sparse Bayesian learning methods can fully exploit the potential prior information in sparse signals. Furthermore, sparse Bayesian learning methods can automatically determine the location and size of nonzero elements in the reconstructed signal through iteration, without requiring prior information on sparsity. Furthermore, sparse Bayesian learning methods can maintain good performance even when the columns of the sensing matrix are highly correlated. Block sparse Bayesian learning (BSBL) is based on sparse Bayesian learning and fully accounts for block correlation in the signal. The sparse signals in unlicensed MIMO-NOMA systems have strong block correlation. Therefore, block sparse Bayesian learning can effectively solve the problem of active user detection and channel estimation in unlicensed MIMO-NOMA systems with human coexistence.

[0005] In block sparse Bayesian learning, the traditional fast marginal likelihood maximization method is often used to optimize the cost function. This method can effectively reconstruct sparse signals in the presence of weak noise. However, when the number of signal blocks is large and the noise is strong, the reconstruction stability and accuracy of this method will be significantly reduced. Therefore, other methods can be considered to optimize the cost function. Q-learning is a model-free reinforcement learning method. During training, it gradually builds a Q-value table, which records the expected reward of taking an action in a certain state. The agent will then choose the action with the highest expected reward in the current state. Using Q-learning to optimize the cost function can improve the reconstruction performance of block sparse Bayesian learning methods.

[0006] The DNN-MP-BSBL method proposed in the prior art is an improvement based on the message-passing block sparse Bayesian method. This method is used in unlicensed MIMO-NOMA systems with a single base station antenna. It shifts the iterative message-passing process from factor graphs to deep neural networks, weights the messages in the DNN, and trains it to minimize estimation error. The Spatiotemporal Structure Enhanced Adaptive Subspace Tracking (STS-ASP) method has also been proposed in the prior art. This method is used in unlicensed MIMO-NOMA systems in human-human coexistence networks. It does not require any prior information (such as the number of active users and noise level) by adaptively obtaining the number of active users and using cross-validation techniques to appropriately terminate the iterative process. This method exploits the spatial correlation between the received signals of different antennas and the temporal correlation between previous and next frames. Furthermore, the Spatially Correlated Block Sparse Bayesian Learning (SC-BSBL) method has been proposed in the prior art. This sparse Bayesian learning method, based on the fast marginal likelihood maximization method, can be used in unlicensed MIMO-NOMA systems in human-human coexistence networks. This method exploits the system's sparsity and spatial correlation and does not require prior information on user sparsity.

[0007] However, existing active user detection and channel estimation methods based on sparse Bayesian classes are all designed for unlicensed MIMO-NOMA systems with a single base station antenna in human-human coexistence networks. In such unlicensed MIMO-NOMA systems, these methods fail to exploit the spatial correlation between different antennas, resulting in poor system performance.

[0008] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0009] The embodiments of the present application provide a method and apparatus for active user detection and channel estimation, to at least solve the technical problem of poor communication performance caused by the inability to utilize the spatial correlation between different antennas.

[0010] According to one aspect of an embodiment of the present application, a method for active user detection and channel estimation is provided, including: determining a channel model and a received signal model in a human coexistence network based on a signal received by a base station; deriving, based on the channel model and the received signal model, the posterior probability density of the channel in the human coexistence network and the total likelihood function of the received signal using the spatial correlation between the channels; obtaining a cost function based on the posterior probability density and the total likelihood function using a type-II maximum likelihood estimation method; and detecting active users in the human coexistence network and estimating the channel based on the cost function.

[0011] According to another aspect of an embodiment of the present application, a joint active user detection and channel estimation device is also provided, including: a model building module, configured to determine a channel model and a received signal model in a human coexistence network based on a signal received by a base station; a derivation module, configured to derive the posterior probability density of the channel in the human coexistence network and the total likelihood function of the received signal based on the channel model and the received signal model and the spatial correlation between the channels; a cost determination module, configured to obtain a cost function based on the posterior probability density and the total likelihood function using a type-II maximum likelihood estimation method; and a detection and estimation module, configured to detect active users in the human coexistence network and estimate the channel based on the cost function.

[0012] In an embodiment of the present application, based on a channel model and a received signal model, the spatial correlation between the channels is utilized to derive the posterior probability density of the channel and the total likelihood function of the received signal in the human coexistence network; based on the posterior probability density and the total likelihood function, a type-II maximum likelihood estimation method is used to obtain a cost function; based on the cost function, active users in the human coexistence network are detected and the channel is estimated; thereby solving the technical problem of poor communication performance caused by the inability to utilize the spatial correlation between different antennas. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0014] Figure 1 It is a flowchart of an active user detection and channel estimation method according to an embodiment of the present application;

[0015] Figure 2 Is a flowchart of another active user detection and channel estimation method according to an embodiment of the present application;

[0016] Figure 3 1 is a schematic structural diagram of an uplink person coexistence network system according to an embodiment of the present application;

[0017] Figure 4 It is a graph showing the change of NMSE with the activation probability according to an embodiment of the present application;

[0018] Figure 5 is a graph of active user detection error rates at different signal-to-noise ratios according to an embodiment of the present application;

[0019] Figure 6 3 is a structural diagram of a joint active user detection and channel estimation device according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0021] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] Example 1

[0023] According to an embodiment of the present application, a method for active user detection and channel estimation is provided, such as Figure 1 As shown, the method includes:

[0024] Step S102, determining a channel model and a received signal model in a human coexistence network based on a signal received by a base station;

[0025] First, determine the expression of the received signal. For example, first obtain the information sequence of each user, and merge the active user indicator symbol into the information sequence to obtain the signal matrix X that merges the active user indicator symbol Lxk ; Then, based on the signal matrix, the channel matrix with block sparse properties, and the noise matrix of the channel, the expression of the signal received by the base station is determined.

[0026] Next, the channel model in the person coexistence network is determined. For example, based on the channel matrix in the person coexistence network The positive definite matrix B representing the correlation result information within the channel m , and the channel matrix The entire block and the received signal matrix The channel model is determined by using the matrix Gaussian probability density expression.

[0027] Then, the signal model in the human coexistence network is determined. For example, based on the signal received by the base station antenna The channel matrix and a positive definite matrix B representing the correlation result information within the channel m , to determine the received signal model.

[0028] Step S104 : Based on the channel model and the received signal model, the spatial correlation between the channels is utilized to derive the posterior probability density of the channel and the total likelihood function of the received signal in the person coexistence network.

[0029] For example, based on the channel model and the received signal model, the Gaussian identity is used to obtain the posterior probability density and the likelihood function of the received signal at each antenna of the base station; the spatial dimension is introduced into the channel model and the received signal model, the received signals on all antennas of the base station are combined into a three-dimensional matrix, and based on the likelihood function of the received signal at each antenna and the three-dimensional matrix, the total likelihood function of all antennas is obtained.

[0030] The posterior probability density function is actually the derived probability of the channel. The larger the probability is, the greater the possibility that the channel is a real channel.

[0031] Existing active user detection and channel estimation methods based on sparse Bayesian classes are all used in unlicensed MIMO-NOMA systems with a single base station antenna in human-human coexistence networks. In such unlicensed MIMO-NOMA systems, these methods fail to exploit the spatial correlation between different antennas, resulting in inferior performance compared to other compressed sensing methods that do account for spatial correlation.

[0032] Existing sparse Bayesian class learning for person coexistence networks suffers from a partial loss of reconstruction performance due to its use of a fast marginal likelihood maximization method, and its reconstruction stability is poor. Furthermore, under low signal-to-noise ratio conditions, existing methods continuously perform block reestimation operations, resulting in poor reconstruction performance. This embodiment addresses these issues in the prior art by leveraging the spatial correlation between channels to derive the posterior probability density of the channels and the total likelihood function of the received signal in the person coexistence network.

[0033] Step S106 : Based on the posterior probability density and the total likelihood function, a cost function is obtained using a type-II maximum likelihood estimation method.

[0034] The cost function is derived from the likelihood function and is used to measure the quality of parameters. Simply put, this embodiment aims to find several parameters that minimize the cost function. This way, the channel and active user estimates based on the parameters that minimize the cost function are the most accurate.

[0035] After obtaining the cost function, the cost function can also be constrained by using the correlation structure within the channel. For example, the correlation coefficient r of the autoregressive model is used to constrain the positive definite matrix in the cost function that represents the correlation result information within the channel. The correlation coefficient r is represented by the basis vector Φ i The sparsity factor s of the degree of overlap between the vectors and the existing vectors in the regression model i , and denote the removal of vector Φ i The quality factor q of the calibrated quantity of the post-model error i To be determined.

[0036] The processed cost function is then optimized using a Q-learning method, which learns reward values based on the current state of the channel and the selected action. For example, the base station, acting as a channel estimation agent, observes the current state and selected action of each channel and, based on the action selected in the current state, determines the reward R for transitioning from the selected action in the current state to the next state. The expected reward for the selected action in the next state is updated based on the reward R for transitioning from the selected action in the current state to the next state, the learning rate alpha, and the discount factor γ used in learning. The cost function is then optimized based on the updated expected reward.

[0037] Some existing joint active user detection methods in human-person coexistence networks rely on spatial correlation between previous and next frame structures. In real-world environments with rapidly changing channel conditions, the prior information from the previous frame can negatively impact channel estimation performance. This embodiment uses Q-learning to solve the cost function, addressing this issue.

[0038] Step S108 : detecting active users in the human coexistence network and estimating channels based on the cost function.

[0039] This embodiment incorporates the spatial dimension when deriving the channel model, received signal model, and total likelihood function. By utilizing the correlation in the spatial dimension, the derived cost function can be optimized to more accurately estimate the channel and active users.

[0040] Example 2

[0041] According to an embodiment of the present application, a joint active user detection and channel estimation method based on spatially correlated block sparse Bayesian learning in a person coexistence network is also provided. In this method, the posterior probability density of the channel and the likelihood function of the received signal are first derived based on expressions for the channel model and the received signal model. The likelihood function is then expressed as a cost function. Finally, the cost function is optimized through Q-learning to obtain the final estimated channel and estimated number of active users.

[0042] Figure 2 The invention relates to a joint active user detection and channel estimation method based on spatially correlated block sparse Bayesian learning in a person coexistence network according to an embodiment of the present application. Figure 2 As shown, the method includes the following steps:

[0043] Step S202: establishing a system model.

[0044] The method of the embodiment of the present application can be applied in a typical uplink human coexistence network, such as Figure 3 As shown in Figure 2, there are a large number of potential users in the human coexistence network, but only a few users are active, while the other users are inactive, that is, the system has the characteristics of sporadic transmission and sparse active users. Suppose there are K users sending data to a central base station, and each user has P a The probability of being active. The system has N subcarriers. Assume that all users are equipped with a single antenna and the base station is equipped with M antennas. In order to meet the needs of massive connections in large-scale Internet of Things, an overload system is considered, that is, N < K. In addition, this embodiment considers a low-density signature OFDM communication system in code domain NOMA and adopts QPSK modulation.

[0045] For each active user, its information sequence is mapped to a sequence of length L c For inactive users, the information sequence is 0. In order to complete channel estimation, each user is assigned a unique training sequence. The length of the training sequence is L s Therefore, each user's information sequence x k The length is L = L c +L s It should be noted that a training sequence that is too long will waste valuable frequency resources. In this embodiment, a Zadoff-Chu (ZC) sequence with good autocorrelation and cross-correlation is used as the training sequence. Then a low-density extended sequence s is used. k x k Each symbol in is mapped to D subcarriers, where the low-density spreading sequence s k is a sparse vector of length N containing D non-zero elements. In this way, the sequence x k Transmitted on D different subcarriers.

[0046] In a human coexistence network, users have the characteristic of sporadic transmission, that is, only a few users are active at the same time. k is defined as the active user indicator of user k. When the user is active, A k Take 1, otherwise A k Take 0. k Merge into the signal matrix, i.e. X L×K =[A1x1,A2x2,…A K x K ] T .

[0047] After that, the signal of the nth carrier received by the base station on the mth antenna can be expressed as:

[0048]

[0049] Among them, s m,n，k is the spreading sequence component of the nth carrier received by user k on the mth antenna. m,n，k is the channel coefficient of the nth carrier received by user k on the mth antenna, which obeys an independent complex Gaussian distribution W m,n is the noise vector, which obeys the complex Gaussian distribution K represents the total number of users.

[0050] At the matrix level, the signal received by the mth antenna at the base station can be expressed as:

[0051]

[0052] Among them, Y m The (l, n)th item of H represents the lth symbol received by the nth subcarrier. The (l, k)th item of X represents the lth symbol sent by the kth user. m The (l, n) term represents h m,n，k =s m,n，k g m,n，k .W m is the noise matrix.

[0053] Then the signal received by M antennas at the base station can be expressed as:

[0054] Y L×N×M =X L×K H K×N×M +W L×N×M (3)

[0055] Among them, X L×K represents the signal matrix sent by all users, W L×N×M represents the total noise matrix.

[0056] In the channel dimension, since user activity is sparse, the channel matrix H is row-sparse. In other words, H has the property of block sparseness. As described in formula (3), H is a three-dimensional matrix of size K×N×M. When there is information on a subcarrier, this means that the channel is a channel vector corresponding to an active user. Otherwise, it is a channel for an inactive user. Therefore, H is sparse in the user dimension. In the carrier dimension, user information is mapped to D subcarriers. The low-density sequence used is sparse, so the carrier dimension is also sparse. In the antenna dimension, the information sent by user k on the same subcarrier n has spatial correlation on different antennas m, and this correlation can be fully utilized.

[0057] In a human coexistence network, the user's activity status and channel information are not a priori information, so it is necessary to obtain relevant information from the received signal. k Combination of, Y can be decomposed into where Y s and Y c Represent the signal matrices of the received training sequence and data sequence respectively. s It can be expressed as:

[0058]

[0059] in, represents the signal matrix of the received training sequence part, Represents the training sequence matrix in the transmitted signal, H K×K×M represents the channel matrix, represents the noise matrix corresponding to the training sequence.

[0060] In the LDS-OFDM system, each user's transmission sequence is mapped to D subcarriers, and the base station knows the prior information of the positions of all non-zero elements. Therefore, the channel matrix H of each user can be eliminated. k The zero elements in are used to simplify the channel matrix H. Formula (4) can be further expressed as:

[0061]

[0062] in, After removing zero matrix, represents the channel matrix after removing zeros, represents the noise matrix after zero removal.

[0063] After decomposition and zero-element removal, the proposed method is used for active user detection and channel estimation. Assuming that the training sequence is known at the base station, the above problem is equivalent to the problem of recovering multidimensional block-sparse signals. In block-sparse Bayesian theory, the most basic mathematical model is:

[0064] Z=ΦA+V (6)

[0065] Where Z is the observation matrix, Φ is the underdetermined perception matrix, A is the block sparse signal matrix to be recovered, and V is the observation noise.

[0066] The essence of the block sparse Bayesian theory is to solve the underdetermined matrix. However, in practical applications, when Φ satisfies the finite isometry, the block sparse signal A can be accurately reconstructed from the observation matrix Z. According to the block sparse Bayesian learning theory and the proposed signal sparse model, the formula (5) It can be regarded as the known perception matrix Φ in formula (6), and the can be regarded as the observation matrix Z in formula (6), and can be viewed as a matrix to be reconstructed. Therefore, the joint user active detection and channel estimation problem in human coexistence networks can be transformed into a multi-dimensional block sparse signal recovery problem.

[0067] Based on the above sparse signal model, this paper proposes a Q-learning-assisted spatially correlated block sparse Bayesian learning (QL-SC-BSBL) method to achieve joint detection of active user detection and channel estimation in human coexistence networks.

[0068] Step S204: Determine the channel model and the received signal model.

[0069] The user's channel has the characteristic of block sparseness. The i-th channel received by the m-th antenna is Obeying complex Gaussian distribution:

[0070]

[0071] Where Γ = diag -1 (γ i ). In order to fully utilize the spatial correlation, the multiple channels between different antennas and the kth user are considered as a whole block, γ i represents the channel matrix The entire block and the received signal matrix For larger γ i , It may be the channel of the active user, and for smaller γ i , It may be noise. i B mi for The covariance matrix of m，i Characterization Channel The correlation structure information within. express Obeying the probability density distribution, According to the matrix Gaussian probability density expression, the channel model at the mth antenna is equivalent to:

[0072]

[0073] in, represents the channel matrix at the mth antenna after removing zeros, T represents, Tr represents the trace of the matrix, B m Represents the correlation structure information within the channel, π KD represents the KD power of π, K represents the total number of users, and D represents the total number of subcarriers.

[0074] For the signal received at the mth antenna It satisfies the following expression:

[0075]

[0076] For the sake of convenience, the definition of Φ in formula (9) is Φ=X s Φ represents the underdetermined perception matrix, β represents the noise variance, Represents the identity matrix of size Ls, where Ls represents the length of the training sequence.

[0077] Step S206 , deriving the posterior probability density of the channel and the likelihood function of the received signal according to the channel model and the received signal model, and obtaining the cost function using the type-II maximum likelihood estimation method.

[0078] According to the channel model and the received signal model, the Gaussian identity can be used to obtain the posterior probability density function at the mth antenna: and likelihood function

[0079]

[0080] in:

[0081]

[0082] ∑ -1 =Γ -1 +βΦ H Φ(13)

[0083]

[0084] Among them, μm represents the estimated mean of the channel at the mth antenna, ∑ represents the estimated variance of the channel at the mth antenna, Φ H represents the conjugate transpose of Φ.

[0085] In order to fully utilize the spatial correlation between the received signals on different receiving antennas, this embodiment combines the received signals on all antennas into a three-dimensional matrix. The total likelihood function of M antennas can be expressed as:

[0086]

[0087] Where M represents the total number of antennas.

[0088] Then, the cost function is obtained using the type-two maximum likelihood estimation method:

[0089]

[0090]

[0091] Step S208: Process the cost function to optimize it.

[0092] The essence of optimizing the cost function (16) is to solve the parameters B, β and γ i , so that the cost function is minimized. In the optimization process, in order to make full use of spatial correlation, the i-th channel on the M antennas is regarded as a complete block signal, that is, the hyperparameter γ i Indicates the correlation of the entire block signal. In practical applications, β is mostly set to a constant. In this embodiment, β is set to The parameter B represents the correlation structure information within the channel H. In practice, the correlation structure information within the channel is similar, so constraints can be added to reduce the computational complexity of the method. This embodiment uses a regression model (AR model) to represent the correlation within the block. When the correlation coefficient of the AR model is r, the matrix B can be constrained by r to B = Toeplitz ([1, r, ... r D-1 ]). Among them, r D-1 Represents r raised to the D-1 power.

[0093] For the parameter γ i The solution of this embodiment is defined as:

[0094]

[0095] The parameter C can be decomposed into:

[0096]

[0097] Here, the subscripts i and j represent the i-th and j-th basis vectors in the matrix, respectively. It represents the remainder after removing the i-th basis vector from C. N represents the total number of vector bases in the parameter C, Φ j represents the jth vector basis in parameter C, Φ i Represents the i-th vector basis in parameter C.

[0098] Then, formula (17) can be further expressed as:

[0099]

[0100] in, s i is the sparse factor, representing the basis vector Φ i The degree of overlap with existing vectors in the model. i is the quality factor, which means that the vector Φ is removed i The calibration amount of the posterior model error.

[0101] Define separately and

[0102]

[0103]

[0104] in, is an independent i function, and Only with γ i In the context of γ i While optimizing, the rest of the basis vectors in the model can be kept constant.

[0105] Step S210: Optimize the processed cost function using Q learning.

[0106] right Q learning is used for optimization, and the settings are as follows:

[0107] Agent: The base station acts as the agent for channel estimation, observing the state and feeding back corresponding rewards.

[0108] Action set: The action set of each estimated channel is determined by the correlation γ i It is defined as A t .

[0109] State set: The state set of each estimated channel is defined as S t ={I t}. t It is a binary variable indicating whether the optimization function is optimizing in a good direction. Less than the previous iteration When I t Takes 1, otherwise takes 0.

[0110] award:

[0111]

[0112] Update of Q value:

[0113] Q t+1 (s, a)←alpha(R+γmaxQ t (s′, a′))+(1-alpha)Q t (s,a) (23)

[0114] Where Q(s, a) is the expected reward for action a in state s. alpha is the learning rate; the smaller it is, the more the agent prioritizes previous training results. γ in Q-learning is the discount factor and is unrelated to the hyperparameters in sparse Bayesian learning. R is the reward for transitioning from the current state to the next state, and t represents the tth update.

[0115] Q learning is to select actions by continuously updating the Q value through the Q table. In each iteration, the parameters ∑, μ and all s i ,q i Perform parameter update. When the iteration is finished, the channel matrix is estimated Estimated active user set where γ T The value can be taken as 0.1 based on experience.

[0116] This embodiment leverages the spatial correlation between received signals from multiple antennas in a human-coexisting network to improve the reconstruction performance of sparse Bayesian learning-based methods. This embodiment establishes a multidimensional signal model in a human-coexisting network and uses a hierarchical sparse Bayesian learning framework to derive a three-dimensional posterior probability density function and likelihood function, as well as a cost function. During the cost function optimization process, the spatial correlation between multiple antennas is exploited to improve the reconstruction performance of this method.

[0117] This embodiment performs channel estimation and active user detection for each frame in a human coexistence network. This embodiment processes received signals within a frame without considering the impact of spatial correlation.

[0118] In addition, this embodiment improves the overall reconstruction performance and stability of the existing likelihood maximization method in the human coexistence network, and solves the problem of poor reconstruction performance under low signal-to-noise ratio conditions. This embodiment uses Q learning to solve the channel matrix in the process of optimizing the cost function. and the received signal matrix The correlation between γi ,While improving the reconstruction performance of the method, the stability of the method reconstruction is also improved, and in the case of low signal-to-noise ratio, the SC-BSBL method is prevented from continuously performing block re-evaluation, which leads to performance degradation.

[0119] Simulation experiment

[0120] This embodiment simulates a frame-based uplink human coexistence network scenario to verify the effectiveness of the QL-SC-BSBL method. In order to simulate a scenario with a large number of users, the number of users K is set to 300. In the simulation, the channel is set to a block fading channel and conforms to a complex Gaussian distribution. The number of antennas M is set to 8, the modulation mode is QPSK modulation, and the active user judgment threshold is set to 0.1. In order to fully compare the impact of various factors on the method, the active user activation probability P a The value range is 0.1 to 0.2, with a step size of 0.02; the system signal-to-noise ratio value range is -5dB to 15dB, with a step size of 5dB.

[0121] The simulation evaluation metrics are the normalized mean square error (NMSE) of the channel and the probability of active user detection error. The probability of active user detection error is the ratio of active users with false detections to the total number of active users. According to the definition, NMSE is expressed as:

[0122]

[0123] Among them, H represents the actual channel, Represents the estimated channel, and NMSE represents the normalized mean square error of the channel.

[0124] For comparison, this example provides the performance of the following four methods:

[0125] (1) Orthogonal Matching Pursuit (OMP): This method is a traditional compressed sensing method that requires prior information about the sparsity of the signal to be reconstructed.

[0126] (2) Generalized Approximate Message Passing Method (GAMP): This method is a derivative of the iterative threshold reconstruction method. It reduces the complexity of the message passing method by approximating the traditional message passing method.

[0127] (3) Random Sparse Learning Multi-User Detection (RSL-MUD): This method treats channel estimation and user activity detection as a dictionary learning problem and solves it based on bilinear generalized approximate message passing.

[0128] (4) Spatially correlated block sparse Bayesian learning method (SC-BSBL): This method derives the cost function based on hierarchical sparse Bayesian learning and optimizes the cost function using the fast marginal likelihood maximization method. The derivation and optimization process fully utilizes the sparse characteristics and spatial correlation of the signal to improve the reconstruction performance of the method.

[0129] Figure 4 is the curve of NMSE changing with activation probability. Figure 4 It can be seen that the NMSE of each method gradually deteriorates as the activation probability increases. The methods used in this simulation are all based on data sparsity. However, as the number of potential users in the system increases, sparsity gradually decreases, causing the NMSE performance of the methods to deteriorate. The proposed QL-SC-BSBL method fully considers the system's block sparsity and the spatial correlation between antennas in the model derivation process, and uses Q-learning to optimize the cost function, which to some extent reduces the impact of increased sparsity on the method's reconstruction performance. Therefore, even in the case of low sparsity, the QL-SC-BSBL method can still maintain a high NMSE level.

[0130] Depend on Figure 4 As can be seen, the OMP method has a poor NMSE and does not vary much with the value of Pa. In simulations, the prior sparsity of the OMP method is set to a fixed value, so the NMSE curve of this method hardly changes with changes in activation probability. The GAMP method reconstructs sparse signals using a message passing factor graph, completing channel estimation without specifying sparsity, and achieves better recovery performance than the OMP method. However, the OMP and GAMP methods do not fully exploit the spatial correlation between receive antennas, which limits their channel estimation performance in MIMO systems. Compared with the above two methods, the RSL-MUD method can better exploit spatial correlation. However, due to its strict requirement for limited isometry of the sensing matrix, the channel estimation performance of the RSL-MUD method is inferior to that of sparse Bayesian learning methods. Due to the use of Q-learning in optimizing the cost function, the QL-SC-BSBL method has better overall performance than the SC-BSBL method, which uses a fast marginal likelihood maximization method. At the same time, from the comparison of the four curves at different signal-to-noise ratios, it can be seen that compared with the QL-SC-BSBL method, the reconstruction performance of the SC-BSBL method decreases more when the signal-to-noise ratio becomes lower.

[0131] Figure 5 is the active user detection error rate of the method under different signal-to-noise ratios when the active user activation probability is 0.1. Figure 3As shown, the active user detection error rate of the proposed method decreases with increasing signal-to-noise ratio (SNR). At low SNRs, the OMP and GAMP methods perform poorly in active user detection, while the other three methods perform significantly better. When the SNR is above 5dB, the SC-BSBL method and the proposed QL-SC-BSBL method perform almost error-free active user detection. Only when a user is active is the corresponding channel activated. Therefore, the performance of active user detection is closely related to the performance of channel estimation. Compared with the other methods, the OMP method has poor channel estimation performance, resulting in poor active user detection performance. Sparse Bayesian methods, based on the block sparse Bayesian principle, detect active users by finding nonzero blocks in sparse data. Therefore, compared with other methods, these methods can always detect active users well when the SNR is low. Because Q-learning is used in optimizing the cost function, the QL-SC-BSBL method avoids the multiple block re-estimations that can occur with the fast marginal likelihood maximization method at low signal-to-noise ratios. Therefore, the QL-SC-BSBL method outperforms the SC-BSBL method in active user detection at low signal-to-noise ratios. However, at high signal-to-noise ratios, the SC-BSBL method does not perform as many block re-estimations, resulting in similar active user detection error probabilities for the two methods.

[0132] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0133] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0134] Example 3

[0135] According to an embodiment of the present application, a joint active user detection and channel estimation device is also provided, such as Figure 6The illustrated example includes a model building module 62 , a derivation module 64 , a cost determination module 66 and a replacement module 68 .

[0136] A model building module 62 is configured to determine a channel model and a received signal model in a human coexistence network based on signals received by the base station;

[0137] A derivation module 64 is configured to derive the posterior probability density of the channel and the total likelihood function of the received signal in the person coexistence network based on the channel model and the received signal model and utilizing the spatial correlation between the channels;

[0138] The cost determination module 66 is configured to obtain a cost function based on the posterior probability density and the total likelihood function using a type II maximum likelihood estimation method;

[0139] The detection and estimation module 68 is configured to detect active users in the human coexistence network and estimate channels based on the cost function.

[0140] Optionally, the specific examples in this embodiment may refer to the examples described in the above-mentioned embodiment 1 and embodiment 2, and this embodiment will not be described in detail here.

[0141] Example 4

[0142] The embodiment of the present application further provides a joint active user detection and channel estimation system, comprising a user equipment and a base station, wherein the base station comprises a joint active user detection and channel estimation device.

[0143] Optionally, the specific examples in this embodiment may refer to the examples described in the above-mentioned embodiment 1 and embodiment 2, and this embodiment will not be described in detail here.

[0144] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0145] If the integrated units in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in the above-mentioned computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling one or more computer devices (which can be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application.

[0146] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0147] In the several embodiments provided in this application, it should be understood that the disclosed client can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, and can be electrical or other forms.

[0148] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0149] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0150] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for active user detection and channel estimation, applied to a license-free MIMO-NOMA system in a human-human coexistence network, characterized in that: include: Determine the channel model and the received signal model in the human coexistence network based on the signal received by the base station; Based on the channel model and the received signal model, the spatial correlation between the channels is utilized to derive the posterior probability density of the channel and the total likelihood function of the received signal in the person coexistence network; Based on the posterior probability density and the total likelihood function, a cost function is obtained using a type-two maximum likelihood estimation method; Based on the cost function, detecting active users in the human coexistence network and estimating channels; Determining the channel model and the received signal model in the human coexistence network includes: based on the channel matrix in the human coexistence network The positive definite matrix B representing the correlation result information within the channel m , and the channel matrix The entire block and the received signal matrix The correlation between the channel model is determined by using the matrix Gaussian probability density expression; based on the signal received by the antenna of the base station The channel matrix and a positive definite matrix B representing the correlation result information within the channel m , determining the received signal model; wherein, based on the channel model and the received signal model, deriving the posterior probability density of the channel and the total likelihood function of the received signal in the human coexistence network, including: based on the channel model and the received signal model, using Gaussian identity to obtain the posterior probability density and the likelihood function of the received signal at each antenna of the base station; introducing the spatial dimension into the channel model and the received signal model, combining the received signals on all antennas of the base station into a three-dimensional matrix, and obtaining the total likelihood function of all antennas based on the likelihood function of the received signal at each antenna and the three-dimensional matrix; Before determining the channel model and the received signal model in the human coexistence network, the method further includes: obtaining an information sequence of each user, and merging an active user indicator symbol into the information sequence to obtain a signal matrix X incorporating the active user indicator symbol. Lxk ; Based on the signal matrix, the channel matrix with block sparse properties, and the noise matrix of the channel, determine the expression of the signal received by the base station.

2. The method according to claim 1, characterized in that After obtaining the cost function using the Type II maximum likelihood estimation method, the method further includes: Using the correlation structure within the channel, adding constraints to the cost function; The processed cost function is optimized using a Q-learning method, wherein the Q-learning method is a method for learning a reward value based on a current state of a channel and a selected action.

3. The method according to claim 2, characterized in that The processed cost function is optimized using a Q-learning method, including: The base station, acting as the channel estimation agent, observes the current state and selected action of each channel, and determines the reward R for transferring the current state to the next state based on the action selected in the current state; Based on the reward R, learning rate, and discount factor γ of the action selected in the current state to transfer to the next state, update the expected reward of the selected action in the next state; The cost function is optimized based on the updated expected reward.

4. The method according to claim 2, characterized in that The process of adding constraints to the cost function by utilizing the correlation structure within the channel includes: The correlation coefficient r of the autoregressive model is used to constrain the positive definite matrix representing the correlation result information within the channel in the cost function.

5. The method according to claim 4, characterized in that The correlation coefficient r is represented by the basis vector Φ i The sparsity factor s of the degree of overlap between the vectors and the existing vectors in the regression model i , and denote the removal of vector Φ i The quality factor q of the calibrated quantity of the post-model error i To be determined.

6. A joint active user detection and channel estimation device, applied to a license-free MIMO-NOMA system in a human coexistence network, characterized in that: include: A model building module is configured to determine a channel model and a received signal model in a human coexistence network based on a signal received by a base station; A derivation module is configured to derive the posterior probability density of the channel and the total likelihood function of the received signal in the person coexistence network based on the channel model and the received signal model and utilizing the spatial correlation between the channels; a cost determination module configured to obtain a cost function using a type-II maximum likelihood estimation method based on the posterior probability density and the total likelihood function; a detection and estimation module configured to detect active users in the human coexistence network and estimate channels based on the cost function; The model building module is further configured to: based on the channel matrix in the person coexistence network The positive definite matrix B representing the correlation result information within the channel m , and the channel matrix The entire block and the received signal matrix The correlation between the channel model is determined by using the matrix Gaussian probability density expression; based on the signal received by the antenna of the base station The channel matrix and a positive definite matrix B representing the correlation result information within the channel m , to determine the received signal model; The derivation module is further configured to: obtain, based on the channel model and the received signal model, the posterior probability density and the likelihood function of the received signal at each antenna of the base station using the Gaussian identity; introduce the spatial dimension into the channel model and the received signal model, combine the received signals on all antennas of the base station into a three-dimensional matrix, and obtain the total likelihood function of all antennas based on the likelihood function of the received signal at each antenna and the three-dimensional matrix; Before determining the channel model and the received signal model in the human coexistence network, the apparatus is further configured to: obtain the information sequence of each user, and merge the active user indicator symbol into the information sequence to obtain a signal matrix X incorporating the active user indicator symbol Lxk ; Based on the signal matrix, the channel matrix with block sparse properties, and the noise matrix of the channel, determine the expression of the signal received by the base station.

7. A human coexistence network, comprising user equipment and a base station, wherein: The base station comprises the joint active user detection and channel estimation apparatus according to claim 6.

Images