Active user detection method, device and computer-readable medium based on decorrelation approximate message passing algorithm

Abstract

The present invention discloses an active user detection method based on a decorrelation approximate message passing algorithm, which relates to the field of signal processing technology. The method comprises: receiving a plurality of leading sequences sent by random access users; executing an approximate message passing algorithm, and iteratively calculating an estimated sequence of each leading sequence from the received signal; in each iterative calculation, using the minimum mean square error of the estimated signal of the previous iteration to design a denoiser, the filter matrix in the denoiser adopts the decorrelation matrix of the leading sequence, and performing a log-likelihood ratio test on the estimated signal to obtain the user's active state detection result. The present invention introduces a decorrelation matrix in the denoiser, so that the estimation error is uncorrelated, and the orthogonality of the error is achieved, thereby improving the accuracy of the detection result in large-scale antenna and high signal-to-noise ratio application scenarios.

Description

Technical Field

[0001] The present invention relates to the field of signal processing technology, and in particular to an active user detection method based on decorrelation approximate message transmission. Technical Background

[0002] As the world's largest wireless network, cellular networks have covered 90% of the world's population and support long-distance, high-throughput, low-latency and highly reliable wireless communications. In particular, in recent years, cellular networks have adopted various emerging wireless transmission access technologies, such as massive multiple-input multiple-output (MIMO), millimeter wave and terahertz technology. Cellular Internet of Things can provide a unified solution for IoT applications in various scenarios. Massive MIMO technology, that is, using large-scale antenna arrays at base stations to provide services for a large number of active devices within the same time and frequency resources through simple signal processing, fundamentally redefines the spatial multiplexing of users and can create a large number of spatial degrees of freedom for users. Massive MIMO achieves two important features: channel hardening and favorable propagation. The occurrence of channel hardening essentially means that large-scale multi-antenna pre-processing or post-processing converts the channel into an almost deterministic amount, making the link quasi-immune to small-scale changes in the instantaneous channel. The occurrence of favorable propagation means that as the size of the antenna array increases, the propagation channels for different users become more separated, and the interference between users is effectively mitigated by using spatial multiplexing technology. In addition, the antenna array gain can improve the signal-to-noise ratio to meet the user's extended coverage requirements.

[0003] With the increasing demand for device connections in cellular IoT, the traditional authorization-based random access technology will lead to long access delays and signaling overhead costs due to its complex signaling transmission process, reducing the utilization efficiency of wireless resources. Therefore, the authorization-based random access protocol is no longer suitable for large-scale machine-type communication scenarios. The introduction of unauthorized random access technology in large-scale connections is the current main research direction. Unauthorized random access technology requires active users to first send pilot sequences to the base station to identify the user identity, and then transmit data signals, allowing user devices to communicate without obtaining authorization from the base station. Since the coherence time of the wireless network channel is limited, it is impossible to assign orthogonal pilot sequences to all potential user devices. Therefore, a challenging problem that needs to be solved urgently is to perform active user detection at the base station by using non-orthogonal pilot sequences, judge the status of potential users, identify active users and perform channel estimation, and then complete the traditional processes such as communication resource scheduling and information transmission for active users.

[0004] In the massive unlicensed random access technology, the base station deploys a large number of antennas through a large-scale MIMO system and performs active user detection based on the non-orthogonal pilot sequence signals sent by the users. This is a typical sparse signal recovery problem. Using the structural characteristics of sparse signals, it can be transformed into the Multiple Measurement Vector (MMV) model in the compressed sensing theory. In recent years, various algorithms have been proposed to solve the problem of active user detection under the MMV model. From the literature search results, the existing research mainly involves the following types of algorithms:

[0005] (1) The sparse Bayesian learning (SBL) algorithm based on non-convex optimization methods is derived from the Bayesian framework. The algorithm is flexible and can adaptively characterize the complete statistical value of the estimated signal based on the prior distribution of different signals. However, due to the complexity of the calculation, it is not suitable for high-dimensional problems.

[0006] (2) The hierarchical block orthogonal matching pursuit (HBOMP) algorithm is an extension of the greedy algorithm based on local optimal selection. The main idea of ​​matching pursuit is to select the atoms that best match the original signal and the residual signal from the dictionary matrix and add them to the recovery set in each iteration, so as to gradually approximate the sparse signal. However, this algorithm requires the sparsity of the signal to be known a priori, and performs least squares estimation through matrix inversion, which has a high computational complexity.

[0007] (3) Approximate Message Passing (AMP) algorithm based on convex optimization problem. This type of algorithm adds the tracking formula of message variables on the basis of iterative shrinkage threshold denoising, so that it has both the low complexity of iterative threshold algorithm and the reconstruction ability of basis pursuit algorithm, further improving the recovery performance of sparse signals. However, it has high requirements on the independent and identical distribution of measurement matrices, otherwise the algorithm itself is prone to divergence.

[0008] In view of the active user detection problem raised in the scenario of massive machine-type communication, the large spatial multiplexing capability and good propagation characteristics of the massive MIMO system can better support multi-user detection for a large number of devices in massive connected communications. However, for the MMV detection problem under the massive MIMO system mechanism, the MMV-AMP algorithm, which designs the minimum mean squared error (MMSE) denoiser based on the statistical information of the wireless channel (large-scale fading coefficient), becomes difficult to converge during the active user detection process, resulting in low accuracy of the results, especially when the base station is configured with a large number of antennas and the received signal-to-noise ratio is large. Therefore, it is necessary to design an active user detection method suitable for scenarios with massive antennas and high signal-to-noise ratios. Summary of the invention

[0009] The present invention provides an active user detection method based on decorrelation approximate message transmission, and related electronic devices and storage media, so as to solve the problem that the detection results of the existing active user detection methods are not accurate in large-scale antenna and high signal-to-noise ratio application scenarios.

[0010] The present invention is achieved through the following technical solutions:

[0011] A first aspect of the present invention provides an active user detection method based on a decorrelation approximate message passing algorithm, which is applied to a base station. The method comprises:

[0012] Receiving a preamble sequence sent by a plurality of random access users to obtain a received signal; the preamble sequence is a non-orthogonal pilot sequence pre-allocated by the base station to the plurality of users;

[0013] Execute an approximate message passing algorithm, iteratively calculate the estimated signal of each leading sequence from the received signal until the iteration stop condition is met to obtain the estimated sequence; wherein, in each iterative calculation, a denoiser is designed using the minimum mean square error of the estimated signal of the previous iteration, the filter matrix in the denoiser is the decorrelation matrix of the leading sequence, and the denoiser is used to perform denoising filtering on the estimated signal of the previous iteration, thereby completing one iterative calculation;

[0014] A log-likelihood ratio test is performed on the estimated signal to obtain a user's activity state detection result.

[0015] The present invention introduces a decorrelation matrix of the leading sequence on the basis of the MMV-AMP detection method, and replaces the filter matrix in the denoiser to perform calculations with the residual signal, so that the estimation error is uncorrelated and the orthogonality of the error is achieved, thereby changing the correlation between the leading sequence and the residual of each iteration, and improving the accuracy of the detection result in large-scale antenna and high signal-to-noise ratio application scenarios.

[0016] Further, iteratively calculating an estimated signal of each preamble sequence from the received signal until an iteration stop condition is met to obtain an estimated sequence, comprising:

[0017] S101, Initialize iterative parameters: Initialize the estimated signal Initialize residual R0 = Y, initialize state variables Initialization error variance Initialize the number of iterations t = 1; where Y is the received signal, M is the number of base station antennas, L is the length of the preamble sequence, and K tot is the number of random access users, σ 2 is the background noise variance at the base station receiving end, ‖R0‖ F is the F norm of the initial residual;

[0018] S102, calculate the decorrelation matrix W of the leading sequence t ;

[0019] S103, based on the estimated signal of user k Design a denoiser using the minimum mean square error of the denoiser, and replace the filter matrix in the denoiser with the decorrelation matrix in S102;

[0020] S104, the estimated signal of user k is processed by the noise reducer in S103 Perform noise reduction filtering to obtain the estimated signal of user k in the tth iteration

[0021] S105, update residual R t , state variable τ t and error variance

[0022] S106, update the number of iterations t=t+1, repeat S102 to S105 until the iteration stop condition is met, and the estimated signal of the last iteration is obtained as

[0023] S107, based on the estimated signal of the last iteration Get the estimated sequence: Among them, W T is the decorrelation matrix calculated at the last iteration, R t is the residual of the last iteration, Right now Yes X T Column vector of .

[0024] Further, the received signal is expressed as:

[0025]

[0026] Among them, α k Indicates the active status of user k. If user k is an active user, then α k =1, otherwise α k =0; k∈K tot , K tot is the total number of random access users; ρ k is the transmission power of user k; s k is the leading sequence sent by user k; h k is the channel vector between user k and the M antennas of the base station; W is the additive Gaussian white noise at the base station receiver, with a mean of 0 and a variance of σ 2 .

[0027] Furthermore, the channel vector h between the user k and the M antennas of the base station is k It is expressed as:

[0028]

[0029] Among them, β k is the large-scale fading value of user k, g km represents the small-scale fading value of user k, M is the total number of antennas of the base station, It means that it obeys Gaussian white noise with mean 0 and variance 1, h k ∈C 1×M , represents a 1×M dimensional vector.

[0030] Furthermore, the decorrelation matrix of the leading sequence is calculated as follows:

[0031] S110, calculating the quasi-decorrelation matrix based on the linear minimum mean square error:

[0032]

[0033] Among them, S is the user's leading sequence matrix, S∈C L×K , is an L×K dimensional matrix; L is the length of the preamble sequence, and K is the total number of users; is the error variance of the tth iteration, S H is the conjugate transpose of S, σ 2 is the background noise variance, I is the unit matrix;

[0034] S120, calculating the decorrelation matrix of the leading sequence based on the quasi decorrelation matrix:

[0035]

[0036] Among them, K tot is the total number of random access users, and tr() is the trace of the matrix.

[0037] Furthermore, the denoiser is expressed as:

[0038]

[0039]

[0040] in, is the preamble sequence estimation signal of user k in the tth iteration, ∈ is the probability of random access for each user, τ t is the state variable of the tth iteration, M is the number of antennas, L is the length of the preamble sequence, and R t is the residual of the tth iteration, ‖R0‖ F is the F-norm of the initial residual, β k is the large-scale fading value of user k, for The conjugate transpose of for The conjugate transpose of .

[0041] Furthermore, the estimated signal is denoised by a denoiser Perform noise reduction filtering to obtain the estimated signal of the tth iteration Specifically:

[0042]

[0043] Among them, η t,k is the denoiser of the tth iteration, W t is the decorrelation matrix of the tth iteration, R t-1 is the residual of the t-1th iteration, is the estimated signal of user k in the t-1th iteration.

[0044] Furthermore, the estimated signal is denoised by a denoiser Perform noise reduction filtering to obtain the estimated signal of the tth iteration Specifically:

[0045]

[0046] Among them, η t,k is the denoiser of the tth iteration, W t-1 is the decorrelation matrix of the t-1th iteration, R t-1 is the residual of the t-1th iteration, is the estimated signal of the t-1th iteration.

[0047] Further, performing a log-likelihood ratio test on the estimated sequence to obtain the user's activity status includes:

[0048] Calculate the decision threshold based on the log-likelihood ratio test The decision rule is expressed as:

[0049]

[0050]

[0051] Among them, LLR represents the log-likelihood ratio test decision rule, then the log-likelihood ratio is less than the decision threshold When , the user is judged to be inactive; the log-likelihood ratio is greater than the decision threshold , the user is judged to be in active state. Indicates that when α k =1, the likelihood probability of the active user is, Indicates that when α k =0, the likelihood probability of the inactive user is, For the judgment threshold;

[0052] According to the decision threshold, the activity state of user k is expressed as follows:

[0053]

[0054] in, is the estimated sequence, yes The conjugate transpose of Indicates that user k is an active user, Indicates that user k is an inactive user.

[0055] A second aspect of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the active user detection method based on the decorrelation approximate message passing algorithm as described in any one of the first aspect of the present invention is implemented.

[0056] According to a third aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the active user detection method based on the decorrelation approximate message passing algorithm described in any one of the first aspects of the present invention is implemented.

[0057] Compared with the prior art, the present invention has the following advantages and beneficial effects: the preamble sequence decorrelation matrix is ​​introduced to replace the filter matrix in the noise reducer, and the estimation error is made uncorrelated by calculating the zone correlation matrix with the residual signal, thereby achieving the orthogonality of the error, thereby changing the correlation between the preamble sequence and the residual of each iteration, and improving the accuracy of the detection result in large-scale antenna and high signal-to-noise ratio application scenarios. Experimental results show that the use of linear minimum mean square error to calculate the decorrelation matrix of the preamble sequence has excellent performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0059] Figure 1 is a flow chart of an active user detection method for decorrelated approximate message delivery according to an embodiment of the present invention;

[0060] Figure 2 is a schematic diagram of an iterative solution process of an embodiment of the present invention;

[0061] Figure 3 is a comparison diagram of active user detection results according to an embodiment of the present invention;

[0062] Figure 4 It is a comparison chart of missed detection rate using three different decorrelation matrices;

[0063] Figure 5 This is a comparison chart of false alarm rates using three different decorrelation matrices. DETAILED DESCRIPTION

[0064] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0065] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to or inherent to other steps or units of the device.

[0066] The terms used in the various embodiments of the application are only used to describe the purpose of specific embodiments and are not intended to limit the various embodiments of the application. As used herein, the singular form is intended to also include the plural form, unless the context clearly indicates otherwise. Unless otherwise limited, all terms used here (including technical terms and scientific terms) have the same meaning as the meanings commonly understood by ordinary technicians in the field of the various embodiments of the application. The terms (such as the terms defined in the dictionary generally used) will be interpreted as having the same meaning as the contextual meaning in the relevant technical field and will not be interpreted as having an idealized meaning or an overly formal meaning, unless clearly defined in the various embodiments of the application.

[0067] The MMV-AMP algorithm exhibits pathological and unpredictable behavior, which affects the normal active user detection performance. In high signal-to-noise ratio areas, the MMV-AMP algorithm has difficulty distinguishing the noise variance of the estimation error from the Gaussian additive white noise at the receiving end, resulting in a tendency to regard inactive users as active. To solve this problem, the present invention designs an active user detection method suitable for large-scale antennas and high signal-to-noise ratio scenarios. By introducing a decorrelation matrix based on the linear minimum mean square error (LMMSE) on the basis of the MMV-AMP algorithm, the stability and robustness of user activity detection are improved, thereby improving the accuracy of the detection results in large-scale antennas and high signal-to-noise ratio application scenarios.

[0068] like Figure 1 As shown, Figure 1 This is a flow chart of an active user detection method based on a decorrelation approximate message passing algorithm of the present invention. The method of the present invention is applied to the base station to detect active users, and then allocate orthogonal preamble sequences to active users to achieve unlicensed communication. It includes the following steps:

[0069] S1, receiving a preamble sequence sent by multiple random access users to obtain a received signal; wherein the preamble sequence is a non-orthogonal pilot sequence pre-allocated by the base station to multiple users;

[0070] S2, executing the approximate message passing algorithm, iteratively calculating the estimated signal of each leading sequence from the received signal until the iteration stop condition is met to obtain the estimated sequence; wherein, in each iterative calculation, the minimum mean square error of the estimated signal of the previous iteration is used to design a denoiser, the filter matrix in the denoiser is the decorrelation matrix of the leading sequence, and the denoiser is used to perform denoising filtering on the estimated signal of the previous iteration, thereby completing one iterative calculation;

[0071] S3, performing a log-likelihood ratio test on the estimated signal to obtain a user's active state detection result.

[0072] The detection environment considers the uplink communication system in the cellular network. The base station antennas are distributed in a large-scale uniform linear array (ULA). Assume that the number of configured antennas is M and K are randomly distributed in the cellular cell. tot single-antenna users, the channel parameters of all users are c The internal remains unchanged and remains independent of each other.

[0073] In step S1, the preamble sequence sent by the user randomly accessing the base station in the cellular cell to the base station is randomly allocated to the users in the cellular cell in advance by the base station, that is, before step S1, the base station has allocated a unique preamble sequence s to each user. k ∈C (L×1) Used for identity authentication. Where k is the user ID, S k represents the leading sequence of user k. The leading sequence used for identity authentication of each user is generated by an independent and identically distributed Gaussian sequence, and the length of each Gaussian sequence is L, satisfying

[0074]

[0075] Each time a user randomly accesses the wireless network, it sends the assigned preamble sequence s to the cellular base station within L time slots. k ; Then the received sequence of the base station can be expressed as a linear combination of the transmitted signal plus the noise term:

[0076]

[0077] Among them, Y is the received signal of the base station, α k Indicates the active status of user k. If user k is an active user, then α k =1, otherwise α k =0; k∈K tot , K tot is the total number of random access users; ρ k is the transmission power of user k; s k K is the leading sequence sent by user k; tot is the total number of random access users; h k is the channel vector between user k and the M antennas of the base station;

[0078] Among them, the channel vector between user k and the M antennas of the base station is expressed as:

[0079]

[0080] in, C 1×M represents a 1×M-dimensional vector; β krepresents the large-scale fading value, g km represents the small-scale fading value of user k, M is the total number of antennas of the base station, is the additive Gaussian white noise at the base station receiving end, with a mean of 0 and a variance of σ 2 , I is the identity matrix.

[0081] The key to detecting user activity is to accurately restore the preamble sequence sent by the user from the received signal. Therefore, in step S2, the base station estimates the preamble sequence sent by each user from the received signal Y through iterative calculation using an approximate message passing algorithm, and then detects user activity based on the estimated preamble sequence.

[0082] Before executing step S2, the base station retrieves the parameters required for estimation from the local storage. The base station pre-stores the parameters required for estimation locally, including the large-scale fading value β from each user to the base station. k , background noise variance σ 2 and user preamble sequence S∈C (L×K) When the base station allocates a non-orthogonal pilot sequence to each user, it stores the allocated pilot sequence locally. Figure 2 The above is a schematic diagram of the iterative solution process. The iterative calculation process in step S2 is specifically as follows:

[0083] S101, Initialize iterative parameters: Initialize the estimated signal Initialize residual R0 = Y, initialize state variables Initialization error variance Initialize the number of iterations t = 1; where Y is the received signal, M is the number of base station antennas, L is the length of the preamble sequence, and K tot is the number of random access users, σ 2 is the background noise variance at the base station receiving end, ‖R0‖ F is the F norm of the initial residual;

[0084] S102, calculate the decorrelation matrix W of the leading sequence t .

[0085] S103, based on the estimated signal of user k The denoiser is designed based on the minimum mean square error and the filter matrix in the denoiser is replaced by the decorrelation matrix in S102.

[0086] S104, estimating the signal through the noise reducer in S103 Perform noise reduction filtering to obtain the estimated signal of the tth iteration

[0087] S105, update residual R t , state variable τ t and error variance

[0088] S106, repeat S102 to S105 until the iteration stop condition is met, and the estimated signal of the last iteration is obtained.

[0089] Furthermore, according to the estimated signal of the last iteration Get the estimated sequence, specifically:

[0090] S107, calculating the estimated sequence according to the estimated signal of the last iteration

[0091]

[0092] Among them, W T is the decorrelation matrix calculated at the last iteration, R T is the residual of the last iteration, Right now Yes X T Column vector of .

[0093] Further, in step S102, the decorrelation matrix of the leading sequence is calculated based on the linear minimum mean square error (LMMSE), including:

[0094] S110, calculating the quasi-decorrelation matrix based on the linear minimum mean square error:

[0095]

[0096] Among them, S is the user's leading sequence matrix, S∈C L×K , is an L×K dimensional matrix; L is the length of the preamble sequence, and K is the total number of access users; is the error variance of the tth iteration, S H is the conjugate transpose of S, σ 2 is the background noise variance, and I is the unit matrix.

[0097] S120, calculating the decorrelation matrix of the leading sequence based on the quasi decorrelation matrix:

[0098]

[0099] Among them, K tot is the total number of random access users, and tr() is the trace of the matrix.

[0100] In step S103, a noise reducer η is designed based on the minimum mean square error (MMES) of the estimated signal estimated by Y. t,k , Indicates that the input signal The signal after filtering is de-noised The denoiser does not change the input signal and can use the large-scale fading statistics of the user to achieve more accurate denoising. By conditional expectation Expands to:

[0101]

[0102]

[0103] Among them, X k is the signal of user k, is the estimated signal of user k, is the preamble sequence estimation signal of user k in the tth iteration, ∈ is the probability of random access for each user, τ t is the state variable of the tth iteration M is the number of antennas, L is the length of the preamble sequence, and R t is the residual of the tth iteration, ‖R0‖ F is the F-norm of the initial residual, β k is the large-scale fading value of user k, for The conjugate transpose of yes The conjugate transpose of .

[0104] In step S104, the estimated signal of user k is Input the denoiser and use the decorrelation function to perform matched filtering on the residual signal of the tth iteration, and the filtered signal is:

[0105]

[0106] Among them, η t,k is the denoiser of the tth iteration, W t is the decorrelation matrix of the tth iteration, R t-1 is the residual of the t-1th iteration, is the estimated signal of user k in the t-1th iteration.

[0107] In step S105, the t-th iteration residual is calculated:

[0108]

[0109] Where Y is the received signal, S is the preamble matrix, and X t-1 is the estimated signal matrix for the t-1th iteration, s k is the leading sequence of user k, (s k ) H Yes k The conjugate transpose of R t-1 is the residual of the t-1th iteration, is the estimated signal of user k at the t-1th iteration, η t,k is the k-th iteration of the denoiser, η′ t,k is η t,k The first derivative of .

[0110] Consider that the mathematical model of the estimated signal in each iteration can be expressed as the sum of the original signal and additive Gaussian white noise, that is:

[0111]

[0112] Among them, z t is a complex Gaussian with zero mean and unit variance, is the estimated signal of user k at the t-th iteration, Error variance in, is the estimated signal, x k is the original signal. Therefore, the error variance of the tth iteration can be obtained by recursive analysis of state evolution:

[0113]

[0114] Among them, η t is the denoising function of the tth iteration, X is the estimated signal, Z t-1 is a complex Gaussian quantity, τ t-1 is the state variable of the t-1th iteration.

[0115] The state variables for the tth iteration are:

[0116]

[0117] in, is the error variance at the tth iteration, σ 2 is the background noise variance, K tot is the total number of users, and L is the length of the preamble sequence.

[0118] Therefore, the error variance at the tth iteration can be expressed as:

[0119]

[0120] In the formula, is the state variable at the tth iteration, σ 2 is the background noise variance, K tot is the total number of users, and L is the length of the preamble sequence.

[0121] In step S106, the number of iterations t=t+1, and steps 102 to S105 are repeatedly executed until the number of iterations reaches the iteration threshold T, or the residual is less than a preset value, then the iteration process is terminated to obtain the estimated signal of the last iteration.

[0122] In step S107, the estimated sequence containing noise is output:

[0123]

[0124] Among them, W T is the decorrelation matrix calculated at the last iteration, R T is the residual of the last iteration, X T is the denoised signal of the last iteration.

[0125] In step S3, the estimated sequence The log-likelihood ratio (LLR) test is used to achieve the optimal decision of user activity results. Specifically, the LLR optimal decision rule is expressed as:

[0126]

[0127] Where LLR represents the log-likelihood ratio test decision rule, Indicates that when α k =1, the likelihood probability of the active user is, Indicates that when α k =0, the likelihood probability of the inactive user is, The judgment threshold.

[0128] According to the likelihood probability of active and inactive users, we can get:

[0129]

[0130] In order to make different users have the same likelihood probability, let The simplified decision expression and threshold are:

[0131]

[0132] Therefore, according to the decision threshold, the activity state of user k is expressed as follows:

[0133]

[0134] in, is the estimated sequence, yes The conjugate transpose of Indicates that user k is an active user, Indicates that user k is an inactive user.

[0135] The present invention detects active users based on the MMV-AMP algorithm of the MMSE denoiser, and adds a decorrelation operation. The decorrelation matrix corresponding to the leading signal sequence matrix S is obtained by using the linear minimum mean square error, and the filter matrix in the denoiser is replaced to calculate with the residual signal, so that the estimation error is uncorrelated, and the orthogonality of the error is achieved, thereby changing the correlation between the leading sequence and the residual of each iteration. Therefore, the method proposed in the present invention is conducive to enhancing the robustness of signal detection of users in different signal-to-noise ratio ranges.

[0136] Based on the above detection method, a simulation experiment is performed on users in a cellular cell to verify the effectiveness of the detection method of the present invention.

[0137] The communication scenario of the simulation experiment is set as follows: the number of base station antennas M = 64, the total number of users in the cellular cell K tot =2000, number of active users K a =120, user transmission power ρ = 15dBm, and other simulation parameter settings are shown in Table 1.

[0138] Table 1 Scenario simulation parameters

[0139]

[0140] The path loss calculated from the above parameters is:

[0141] β k [dB] = -128.1-37.6log 10 (d k )

[0142] Where dk is the horizontal distance from user k to the base station, in km, d k The value of is randomly selected based on the middle distance [0.2km, 0.5km] of the user distribution.

[0143] The Enhanced-AMP algorithm using damping and adaptive iteration, and the AMP algorithm are compared with the detection algorithm of the present invention, such as Figure 3 The figure shows a comparison chart of active user detection results. In the figure, the horizontal axis is the leading sequence length, and the vertical axis is the active user detection error rate. The error detection rate = missed detection rate + false alarm rate. The missed detection rate is the ratio of the detection results that judge active users as inactive users to the total detection results. The false alarm rate is the ratio of the detection results that judge inactive users as active users to the total detection results.

[0144] from Figure 3 It can be seen that when the preamble length is less than 105, the active user detection error probability of the Enhanced-AMP method is optimal, but the performance tends to be flat as the preamble length increases; when the preamble length is greater than 110, the active user detection error probability of the method based on the present invention decreases sharply as the input preamble length increases, which is significantly different from the other two AMP methods. Therefore, when the preamble length is greater than the number of active users, the output performance of the method of the present invention is the best.

[0145] Figure 4 , Figure 5 The following is a comparison of the detection results of using three different decorrelation matrices to replace the filter matrix of the denoiser. Figure 4 This is a comparison chart of missed detection rates using three different decorrelation matrices. Figure 5 The figure below is a comparison of false alarm rates using three different decorrelation matrices. The three decorrelation matrices are LMMSE decorrelation matrix, pseudo-inverse (PINV) decorrelation matrix, and matched filter (MF) decorrelation matrix. Figure 4 and Figure 5 It can be seen that the active user detection error rate based on the MF decorrelation method tends to be stable as the length of the leading sequence increases; the active user detection error rate based on the PINV decorrelation method drops sharply as the length of the leading sequence increases, but there is still a gap with the performance based on LMMSE; therefore, the output performance of the LMMSE decorrelation method is the best, ensuring the minimization of the estimation error.

[0146] According to a second aspect of the present invention, an electronic device is provided, which includes a processor, a memory, an input device, an output device and a communication device; the number of processors in the computer device can be one or more, and the processor, memory, input device and output device in the electronic device can be connected via a bus or other means.

[0147] The memory, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules. The processor executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory, so as to implement the active user detection method based on the decorrelation approximate message passing algorithm of any of the above embodiments of the present invention.

[0148] The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0149] In a third aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the active user detection method based on the decorrelation approximate message passing algorithm of any of the above embodiments of the present invention is implemented. The storage medium may be a ROM / RAM, a magnetic disk, an optical disk, etc.

[0150] A fifth aspect of the present invention provides a computer program product. When the computer program product runs on a computer, it enables the computer to execute the active user detection method based on the decorrelation approximate message passing algorithm of any of the above embodiments of the present invention.

[0151] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An active user detection method based on a decorrelation approximate message passing algorithm, applied to a base station, characterized in that: The method comprises: Receiving a preamble sequence sent by a plurality of random access users to obtain a received signal; the preamble sequence is a non-orthogonal pilot sequence pre-allocated by the base station to the plurality of users; Execute an approximate message passing algorithm, iteratively calculate the estimated signal of each leading sequence from the received signal until the iteration stop condition is met to obtain the estimated sequence; wherein, in each iterative calculation, a denoiser is designed using the minimum mean square error of the estimated signal of the previous iteration, the filter matrix in the denoiser is the decorrelation matrix of the leading sequence, and the denoiser is used to perform denoising filtering on the estimated signal of the previous iteration, thereby completing one iterative calculation; Performing a log-likelihood ratio test on the estimated sequence to obtain an activity status detection result of the user; Iteratively calculating an estimated signal of each preamble sequence from the received signal until an iteration stop condition is satisfied to obtain an estimated sequence, comprising: S101, Initialize iterative parameters: Initialize the estimated signal , initialize the residual , initialize the state variables , initialization error variance , initialize the number of iterations ;in, Y To receive the signal, M is the number of base station antennas, L is the length of the leading sequence, is the number of random access users, is the background noise variance at the base station receiving end, is the F norm of the initial residual; S102, calculating the decorrelation matrix of the leading sequence ; S103, based on user k The estimated signal Design a denoiser using the minimum mean square error of the denoiser, and replace the filter matrix in the denoiser with the decorrelation matrix in S102; S104, using the noise reducer in S103 to k The estimated signal Perform noise reduction filtering to obtain the t Users of iteration k The estimated signal ; S105, update residual , state variables and error variance ; S106, update the number of iterations , repeat S102~S105 until the iteration stop condition is met, and the estimated signal of the last iteration is ; S107, based on the estimated signal of the last iteration Get the estimated sequence: ,in, is the decorrelation matrix calculated at the last iteration, is the residual of the last iteration, ,Right now yes Column vector of .

2. The active user detection method based on decorrelation approximate message passing algorithm according to claim 1 is characterized in that: The received signal is expressed as: in, Indicates user k Active status, if the user k For active users, ,on the contrary ; , is the total number of random access users; For users k The transmission power; For users k The preamble sequence sent; For users k With base station M The channel vector between the root antennas; W is the additive Gaussian white noise at the base station receiving end, with a mean of 0 and a variance of .

3. The active user detection method based on decorrelation approximate message passing algorithm according to claim 2 is characterized in that: The user k With base station M The channel vector between the root antennas It is expressed as: in, For users k The large-scale fading value of Indicates user k The small-scale fading value of, M is the total number of antennas of the base station, , which means it obeys Gaussian white noise with mean 0 and variance 1. ,express dimensional vector.

4. The active user detection method based on decorrelation approximate message passing algorithm according to claim 1 is characterized in that: The decorrelation matrix of the leading sequence is calculated as follows: S110, calculating the quasi-decorrelation matrix based on the linear minimum mean square error: in, is the user's leading sequence matrix, ,for Dimensional matrix; L is the length of the leading sequence, is the total number of users; For the t The error variance of the iteration, for The conjugate transpose of is the background noise variance, is the identity matrix; S120, calculating the decorrelation matrix of the leading sequence based on the quasi decorrelation matrix: in, is the total number of random access users, is the trace of the matrix.

5. The active user detection method based on decorrelation approximate message passing algorithm according to claim 1, characterized in that: The denoiser is represented as: in, It is t Secondary users k The leading sequence estimation signal, is the probability of random access for each user, For the t The state variables of the iteration, , M is the number of antennas, is the length of the leading sequence, For the t The residual error of the iteration, For the t The F-norm of the residual error of the iteration, For users k The large-scale fading value of for The conjugate transpose of for The conjugate transpose of .

6. The active user detection method based on decorrelation approximate message passing algorithm according to claim 1, characterized in that: The estimated signal is processed by the noise reducer Perform noise reduction filtering to obtain the t The estimated signal of the iteration , specifically: in, For the t Iterative denoiser, For the The decorrelation matrix of the iteration, For the The residual error of the iteration, For the Users of iteration k The estimated signal.

7. The active user detection method based on decorrelation approximate message passing algorithm according to claim 1, characterized in that: The performing a log-likelihood ratio test on the estimated sequence to obtain the user's activity status includes: Calculate the decision threshold based on the log-likelihood ratio test , the decision rule is expressed as: in, represents the log-likelihood ratio test decision rule, then the log-likelihood ratio is less than the decision threshold When , the user is judged to be inactive; the log-likelihood ratio is greater than the decision threshold , the user is judged to be in active state; Indicates when When , the likelihood probability of the active user is, Indicates when When , the likelihood probability of the inactive user is, For the judgment threshold; According to the judgment threshold, the user The activity status is represented as follows: in, is the estimated sequence, yes The conjugate transpose of Indicates user For active users, Indicates user Inactive users.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the active user detection method based on the decorrelation approximate message passing algorithm according to any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the active user detection method based on the decorrelation approximate message passing algorithm according to any one of claims 1 to 7 is implemented.

Images