Active user detection system and method based on access point clustering

Abstract

The invention discloses an active user detection system and method based on access point clustering, and belongs to the technical field of active user detection methods. The detection method comprises the following steps: S1, covariance modeling and parameter estimation: constructing a covariance matrix containing a multi-dimensional channel parameter through a signal received by an access point, defining a product of an active state and transmitting power as an active parameter, and estimating a user active state based on the covariance matrix and the active parameter; s2, spatial clustering optimization: performing spatial clustering on the access points by using a K-means clustering algorithm, and optimizing calculation complexity; and S3, active parameter iterative optimization: optimizing active parameters in the clustered covariance matrix by using a coordinate descent method, and determining an optimal solution of the active parameters. Compared with the prior art, the method has the advantages that the system complexity can be remarkably reduced while the detection precision is ensured, and the communication requirements of 6G low power consumption and high-density access in the future are met.

Description

Technical Field

[0001] The present invention relates to the technical field of active user detection methods, and in particular to an active user detection system and method based on access point clustering. Background Art

[0002] In massive machine-type communications (mMTC) scenarios, due to the large number of connected devices and the sudden and intermittent nature of communication, accurate detection of active users is a key technical step in ensuring efficient allocation of network resources and improving the quality of communication services. Currently, active user detection technologies for this scenario are mainly divided into two main approaches: compressed sensing algorithms and covariance-based algorithms.

[0003] Active user detection technology based on compressed sensing leverages the sparse nature of device activity, cleverly transforming the active user identification problem into a sparse reconstruction problem, which is then solved using compressed sensing technology based on a greedy algorithm. This method effectively detects active users through compressed signal sampling and reconstruction. Active user detection technology based on the covariance algorithm involves the base station performing in-depth analysis of signals received from multiple access points, leveraging the inherent properties of the covariance matrix to determine the operating status of active devices. Compared to compressed sensing, the covariance algorithm leverages the advantages of multi-antenna technology, significantly improving detection accuracy. It also eliminates the need for prior knowledge of the number of active users, demonstrating strong adaptability in high-density scenarios. Furthermore, the active user detection method based on primary access point clustering, an extension of the covariance algorithm, involves the user device selecting the top few primary connection access points based on signal quality. Only these key access points are used for active user detection calculations, effectively avoiding redundant calculations for access points with poor signal quality.

[0004] However, existing detection algorithms still have many technical bottlenecks that need to be solved in practical applications. Detection algorithms based on compressed sensing face the problem of rapidly increasing computational complexity in complex communication environments with a large number of devices and drastic signal dynamic changes. When deployed in large networks, their time cost and storage resource consumption are difficult to meet actual needs; when the number of active users exceeds the length of the signal sequence, the algorithm's detection performance will degrade significantly. Covariance-based detection algorithms need to calculate the large-scale fading coefficients of all access points associated with a single user, which triggers large-scale matrix operations and leads to high computational complexity. The covariance detection algorithm based on the first few clusters has application limitations and fails to fully integrate the information of all access points within the user's coverage area. This not only causes unnecessary waste of computing resources, but also fails to achieve optimal control of the probability of detection errors.

[0005] In view of this, the present invention proposes an active user detection method based on access point clustering. Summary of the Invention

[0006] The object of the present invention is to provide an active user detection system and method based on access point clustering to solve the problems mentioned in the background technology.

[0007] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0008] An active user detection system based on access point clustering includes the following modules:

[0009] K-means network function module: This module uses the K-means algorithm to cluster the geographic locations of access points, calculates the distance between users and access points, and assigns access points to different clusters based on distance. This module not only reduces computational complexity but also provides input data for subsequent large-scale fading coefficient calculations and active user detection.

[0010] Covariance Matrix Optimization and Activity Mode Decoding Function Module: This module decodes active user patterns and calculates user activity parameters based on received signals. By minimizing the negative log-likelihood function and optimizing the covariance matrix, it gradually adjusts each user's activity state parameters to accurately determine whether the user is active.

[0011] Large-scale fading coefficient calculation module: This module calculates the large-scale fading coefficient within each cluster based on the distance between the user and the access point, and uses a weighted average method to obtain the overall large-scale fading coefficient for each cluster. This process helps the model more accurately assess the signal strength of different users and optimize subsequent active user detection.

[0012] Signature sequence generation module: Generates a unique signature sequence for identifying each user. By generating a signature sequence with a complex Gaussian distribution, it provides a unique signal identifier for the user device to ensure the accuracy of signal decoding and activity status detection.

[0013] Complex Gaussian noise generation module: This module simulates noise in wireless channels, generates complex Gaussian noise, and adds it to the received signal. This allows the system to simulate a real-world wireless communication environment and help evaluate active user detection performance under different signal-to-noise ratios.

[0014] Missed Detection and False Alarm Probability Calculation Module: This module calculates the missed detection probability and false alarm probability, two key performance indicators of the active user detection system, and adjusts the detection results based on different thresholds. This module allows the system to evaluate detection performance under different threshold settings, thereby optimizing the accuracy of active user detection.

[0015] Constant parameter calculation module: This module is responsible for calculating the polynomial coefficients used to decode the user's activity status. By calculating the polynomial coefficients related to the large-scale fading coefficient, it optimizes the system's decoding process and provides mathematical support for accurate estimation of the user's activity status.

[0016] Main program module: This module combines the functions of the aforementioned modules for overall simulation and calculation, performs Monte Carlo simulations to evaluate detection performance under different signal-to-noise ratios and active user probabilities, and calculates the system's false alarm and missed detection probabilities by integrating large-scale fading coefficients, cluster allocation, and active state estimation results, and outputs performance evaluation graphs.

[0017] An active user detection method based on access point clustering implemented by the system includes the following steps:

[0018] S1. Covariance Modeling and Parameter Estimation: A covariance matrix containing multidimensional channel parameters is constructed from the signals received by the access point. The activity parameter is defined as the product of the activity state and the transmit power. The user activity state is estimated based on the covariance matrix and the activity parameter.

[0019] S2, spatial clustering optimization: Use the K-means clustering algorithm to spatially cluster the access points and optimize the computational complexity;

[0020] S3. Iterative optimization of active parameters: Use the coordinate descent method to optimize the active parameters in the clustered covariance matrix to determine the optimal solution for the active parameters.

[0021] Preferably, the signal in S1 includes active state, transmit power, large-scale fading coefficient, non-orthogonal pilot sequence and noise matrix.

[0022] Preferably, the S1 specifically includes the following contents:

[0023] S1.1. After receiving the signal through the access point and modeling the channel, each access point calculates the signal covariance matrix based on a probabilistic algorithm.

[0024] S1.2. Define the product of the activity state and the transmit power as an activity parameter. Set a threshold. When the estimated activity parameter is greater than the threshold, the system is considered active. When it is less than the threshold, the system is considered inactive.

[0025] S1.3. Use the covariance matrix to obtain the maximum likelihood estimation function for the activity parameter. Estimate the active user status by maximizing the maximum likelihood estimation function, and find the activity parameter that makes user activity most likely to occur in the network.

[0026] S1.4. Establish the negative logarithm of the minimum maximum likelihood estimation function for subsequent minimization optimization to ensure the maximization of probability while simplifying the optimization process.

[0027] Preferably, the S2 specifically includes the following contents:

[0028] S2.1. Based on the K-means clustering algorithm, after specifying the number of clusters, the K-means internal iteration is used to find the optimal cluster center under the current number of access points, and the geographical location is used to plan the access point clusters;

[0029] S2.2. After clustering is completed, calculate the weighted average large-scale fading coefficient of all access points in the cluster;

[0030] S2.3. Select the cluster with the largest weighted average large-scale fading coefficient for the user, establish a connection with it, and perform active user detection modeling according to the covariance algorithm. Replace the original large-scale fading coefficient with the weighted average large-scale fading coefficient and minimize the negative logarithm of the maximum likelihood estimation function to achieve active user detection.

[0031] Preferably, the S3 specifically includes the following contents:

[0032] The negative logarithm of the maximum likelihood function containing the covariance matrix is constructed as the cost function in the coordinate descent method. In each iteration, an active parameter is selected for optimization while keeping other parameters unchanged. The active parameter optimization process includes:

[0033] Calculate the derivative of the active parameter and set its derivative to 0 to find the optimal value that minimizes the cost function;

[0034] All active parameters are gradually adjusted and updated until all active parameters converge and the objective function no longer changes significantly. Finally, the active parameters that maximize the maximum likelihood function are obtained, which is defined as the optimal solution.

[0035] The present invention further protects a computer device, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, and the instruction, program, code set, or instruction set is loaded and executed by the processor to implement the above-mentioned active user detection method based on access point clustering.

[0036] The present invention further protects a computer-readable storage medium, characterized in that the computer-readable storage medium stores at least one instruction, at least one program, code set or instruction set, and the instruction, program, code set or instruction set is loaded and executed by a processor to implement the above-mentioned active user detection method based on access point clustering.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] (1) Traditional active user detection methods based on compressed sensing and covariance algorithms typically require processing complex signal models and large amounts of data, which can lead to excessive computational complexity, especially in scenarios with multiple users and large-scale device access. This invention spatially clusters access points, aggregating multiple access points into several clusters. This significantly reduces computational complexity and improves detection efficiency.

[0039] (2) By clustering access points based on the K-means algorithm, the system proposed in the present invention can improve signal resolution when processing large-scale device access and reduce computational redundancy caused by an excessive number of access points. At the same time, the clustering method can effectively reduce the probability of missed detection and false alarm, thereby improving overall detection performance. Compared with traditional methods, especially under asynchronous conditions, this method can more accurately distinguish between active and inactive users by optimizing intra-cluster collaboration, thereby reducing the overall computational efficiency of the system while ensuring the accuracy of active user detection, and has greater robustness.

[0040] (3) The present invention can flexibly adapt to the distribution of user devices and access points in decellularized large-scale multi-input multi-output networks and has strong scalability. As the number of user devices increases, the K-means clustering algorithm can dynamically adjust the size and number of clusters, ensuring stable detection performance without increasing excessive computational burden.

[0041] (4) The K-means algorithm only needs to perform a single intra-cluster calculation at the beginning, thus reducing computational and storage overhead. In traditional detection methods based on clustering of primary access points, each access point needs to independently calculate the large-scale fading coefficient with the user, which is inefficient in large-scale networks. The present invention reduces repeated computations by sharing computation results among access points within the cluster, saving storage space and computation time.

[0042] In summary, the present invention provides an active user detection system and method based on K-means clustering and covariance matrix optimization, which is suitable for machine-type communication scenarios in de-cellularized large-scale multi-input and multi-output networks. It can significantly reduce system complexity while ensuring detection accuracy, and is more adaptable to the future 6G low-power and high-density access communication requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction to the drawings involved in the embodiments is now provided. It is obvious that the drawings described below are only schematic illustrations of some embodiments of the present invention. Those skilled in the art can construct other forms of drawings based on these drawings without inventive effort.

[0044] Figure 1This is a flowchart of the active user detection method based on access point clustering proposed in Example 1 of the present invention;

[0045] Figure 2 This is a schematic diagram of the computational complexity of the simulation verification proposed in Example 1 of the present invention.

[0046] Figure 3 Schematic diagram of an active user detection system based on K-means clustering and covariance proposed in Example 2 of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments.

[0048] The active user detection system and method based on access point clustering proposed by the present invention will be described below with reference to relevant drawings and specific examples.

[0049] The specific contents are as follows.

[0050] Example 1

[0051] See also Figure 1 The present invention proposes an active user detection method based on access point clustering, which specifically includes:

[0052] S1. Build a decellularized massive MIMO system for active user detection, consisting of multiple access points and users. Assume that the decellularized massive MIMO system has M access points equipped with N antennas, serving K randomly distributed single-antenna users.

[0053] In step S1, due to the sporadic nature of IoT services, only a small number of users in the system are active, while the majority of the remaining users are idle. The probability of a user being active in the system is called the activation probability ε<<1. Let x k ∈{0,1}, when x k =1 indicates that the kth single-antenna user is active; when x k = 0, indicating that the kth single-antenna user is inactive, and let Pr(x k =1)=ε,Pr(x k =0)=1-ε. Let x=(x1,x2,L,x K ) represents the user activity state vector of K single-antenna users at any time. Since mMTC traffic is sporadic, x is sparse and the active user set is represented by the set Assume m = 1, 2, ..., M, n = 1, 2, ..., N, k = 1, 2, ..., K, then the channel gain between the nth antenna in the mth access point and the kth single-antenna user is:

[0054]

[0055] Among them, g mnk is the small-scale fading coefficient that obeys the standard normal distribution, β mk represents the large-scale fading coefficient between the mth access point and the kth user, which includes path loss and shadow fading and is known at the central processing unit.

[0056] S2. After the communication channel is modeled, the signal vector received at each access point is composed of the noise generated during the communication process and the superposition signal of the user signal. The signal matrix received at the mth AP for:

[0057]

[0058] Among them, the matrix is the set of all signature sequences, D x =diag(x) represents the diagonal matrix of the active matrix, represents the diagonal matrix of transmission power, is the channel matrix between the kth user and the mth AP, is the noise matrix. The signal Y received by each access point m Construct a covariance matrix containing multidimensional channel parameters

[0059] S3. Use the K-means clustering algorithm to spatially cluster the access points and optimize the computational complexity. Specifically, the following steps are performed:

[0060] S3.1. Based on the K-means clustering algorithm, after specifying the number of clusters C, use K-means internal iteration to find the optimal cluster center for the current number of access points, and use geographic location to plan access point clusters;

[0061] S3.2. After clustering is completed, the weighted average large-scale fading coefficient of all access points in cluster c = 1, 2, ..., C is calculated and can be expressed as:

[0062]

[0063] in, Refers to the set of access points contained in the cth cluster, d mkis the distance between access point m and user k, and λ is a very small number to prevent division by zero errors. Next, for each user device k, it selects the access point cluster with the largest large-scale fading coefficient. This connection method is called C-β in this invention. max C-β max It can be expressed as:

[0064]

[0065] S4. The user selects the cluster with the largest weighted average large-scale fading coefficient and establishes a connection with it. k and transmit power The product of is defined as the active parameter γ k , the active parameters of each user in the system form the joint active parameter vector The maximum likelihood estimation function of the active parameters is obtained using the covariance matrix, and the negative logarithm of the maximum likelihood estimation function is minimized. The coordinate descent method is used to optimize the active parameters in the clustered covariance matrix to determine the optimal solution for the active parameters and implement active user detection for the system. S4 specifically includes the following:

[0066] S4.1. Using the covariance matrix Σ m The block diagonal structure of , the likelihood estimate of the joint activity parameter γ is:

[0067]

[0068] S4.2. Maximum likelihood estimation of γ * It can be obtained by maximizing p(Y|γ) or minimizing -log(p(Y|γ)). * In the process of active user detection, the coordinate descent algorithm is used. The cost function in the coordinate descent method is constructed by taking the negative logarithm of the maximum likelihood function containing the covariance matrix. In each iteration, an active parameter is selected for optimization, while the other parameters remain unchanged. The active parameter optimization process includes: calculating the derivative of the active parameter, setting it to 0, and finding the optimal value that minimizes the cost function; gradually adjusting and updating all active parameters until all active parameters converge and the objective function no longer changes significantly. Finally, the active parameter that maximizes the maximum likelihood function is obtained, which is defined as the optimal solution.

[0069] S4.3. Obtain the activity parameter γ of each user after the final iteration k After that, we will conduct active user detection and performance evaluation. When conducting active user detection, the final value of each user's active parameter iteration is The threshold k for each device is required For comparison:

[0070]

[0071] Devices whose active user parameter is greater than a threshold are marked as active, and devices whose active user parameter is less than the threshold are marked as inactive.

[0072] S4.4. The receiver operating characteristic (ROC) metrics used to evaluate performance include missed detection probability and false alarm probability, which are important metrics for active user detection. The missed detection probability and false alarm probability are:

[0073]

[0074] in, Represents the actual set of active user devices, Represents the estimated set of active user devices. Design a simulation experiment to verify the time difference between active user detection based on primary access point clustering and active user detection based on K-means clustering. Figure 2 ,Depend on Figure 2 As can be seen, the active user detection time based on K-means clustering is much shorter than that based on master access point clustering. This is due to differences in the nature of the algorithms. The active user detection algorithm based on master access point clustering calculates the distance from each access point to the user and selects the master access point each time. This operation is inefficient in large-scale networks. In contrast, the active user detection algorithm based on K-means clustering divides the access points into several clusters. The access points within each cluster share some calculation results, reducing repeated calculations and thus reducing the amount of data and complexity in subsequent calculations. Secondly, due to the difference in computational complexity, the algorithm based on master access points calculates the user's large-scale fading coefficient separately for each access point, which may introduce more computational overhead. In contrast, the K-means algorithm uses the center of the cluster to represent the large-scale fading coefficients of all access points in the cluster and performs a weighted average calculation. This significantly reduces the number of calculations. At the same time, algorithms based on master access points require more storage and processing of matrix operations because each access point and each user must be processed separately. The K-means algorithm, on the other hand, reduces the complexity of matrix operations by reducing the number of effective access points, saving storage space and computation time. Finally, in a parallel computing environment, algorithms based on master access points, due to their need for global information, may be subject to increased synchronization and data transmission overhead during parallel computing. However, the K-means algorithm, due to its relatively uniform and independent computational tasks, can more efficiently allocate computing resources.

[0075] Example 2:

[0076] Based on Example 1, the present invention provides an active user detection system based on K-means clustering and covariance matrix optimization, see Figure 3,include:

[0077] K-means network function module: This module uses the K-means algorithm to cluster access point locations, calculates the distance between users and access points, and assigns access points to different clusters based on distance. This module not only reduces computational complexity but also provides input data for subsequent large-scale fading coefficient calculations and active user detection.

[0078] Covariance Matrix Optimization and Activity Mode Decoding Function Module: This module decodes active user patterns and calculates user activity parameters based on received signals. By minimizing the negative log-likelihood function and optimizing the covariance matrix, it gradually adjusts each user's activity state parameters to accurately determine whether the user is active.

[0079] Large-scale fading coefficient calculation module: This module calculates the large-scale fading coefficient within each cluster based on the distance between the user and the access point, and uses a weighted average method to obtain the overall large-scale fading coefficient for each cluster. This process helps the model more accurately assess the signal strength of different users and optimize subsequent active user detection.

[0080] Signature sequence generation module: Generates a unique signature sequence for identifying each user. By generating a signature sequence with a complex Gaussian distribution, it provides a unique signal identifier for the user device to ensure the accuracy of signal decoding and activity status detection.

[0081] Complex Gaussian noise generation module: This module simulates noise in wireless channels, generates complex Gaussian noise, and adds it to the received signal. This allows the system to simulate a real-world wireless communication environment and help evaluate active user detection performance under different signal-to-noise ratios.

[0082] Constant parameter calculation module: This module is responsible for calculating the polynomial coefficients used to decode the user's activity status. By calculating the polynomial coefficients related to the large-scale fading coefficient, it optimizes the system's decoding process and provides mathematical support for accurate estimation of the user's activity status.

[0083] Missed Detection and False Alarm Probability Calculation Module: This module calculates the missed detection probability and false alarm probability, two key performance indicators of the active user detection system, and adjusts the detection results based on different thresholds. This module allows the system to evaluate detection performance under different threshold settings, thereby optimizing the accuracy of active user detection.

[0084] Main program module: This module combines the functions of the aforementioned modules for overall simulation and calculation, performs Monte Carlo simulations to evaluate detection performance under different signal-to-noise ratios and active user probabilities, and calculates the system's false alarm and missed detection probabilities by integrating large-scale fading coefficients, cluster allocation, and active state estimation results, and outputs performance evaluation graphs.

[0085] It should be noted that, in the present invention patent, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0086] The above are only preferred specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. An active user detection system based on access point clustering, characterized in that: Includes the following modules: K-means network function module: Uses the K-means algorithm to cluster the geographic locations of access points, calculates the distance between users and access points, and assigns access points to different clusters based on the distance; Covariance matrix optimization and active mode decoding function module: used to decode active user patterns and calculate user activity parameters based on received signals; By minimizing the negative log-likelihood function and optimizing the covariance matrix, the activity status parameters of each user are gradually adjusted to accurately determine whether the user is active; Large-scale fading coefficient calculation module: Calculates the large-scale fading coefficient within each cluster based on the distance between the user and the access point, and obtains the overall large-scale fading coefficient of each cluster through the weighted average method; Signature sequence generation module: Generates a unique signature sequence for identifying each user. By generating a signature sequence with a complex Gaussian distribution, it provides a unique signal identifier for the user device to ensure the accuracy of signal decoding and activity status detection. Complex Gaussian noise generation module: used to simulate the noise in the wireless channel, generate complex Gaussian noise and add it to the received signal; Missed detection and false alarm probability calculation module: used to calculate the missed detection probability and false alarm probability, two key performance indicators of the active user detection system, and adjust the detection results according to different thresholds; Constant parameter calculation module: responsible for calculating polynomial coefficients for decoding user activity status; Main program module: This module combines the functions of the aforementioned modules for overall simulation and calculation, performs Monte Carlo simulations to evaluate detection performance under different signal-to-noise ratios and active user probabilities, and calculates the system's false alarm and missed detection probabilities by integrating large-scale fading coefficients, cluster allocation, and active state estimation results, and outputs performance evaluation graphs.

2. A method for detecting active users based on access point clustering implemented by the system of claim 1, characterized in that: The following steps are involved: S1. Covariance Modeling and Parameter Estimation: A covariance matrix containing multidimensional channel parameters is constructed from the signals received by the access point. The activity parameter is defined as the product of the activity state and the transmit power. The user activity state is estimated based on the covariance matrix and the activity parameter. S2, spatial clustering optimization: Use the K-means clustering algorithm to spatially cluster the access points and optimize the computational complexity; S3. Iterative optimization of active parameters: Use the coordinate descent method to optimize the active parameters in the clustered covariance matrix to determine the optimal solution for the active parameters.

3. The method for detecting active users based on access point clustering according to claim 2, wherein: The signal in S1 includes active state, transmit power, large-scale fading coefficient, non-orthogonal pilot sequence and noise matrix.

4. The method for detecting active users based on access point clustering according to claim 3, wherein: The S1 specifically includes the following contents: S1.

1. After receiving the signal through the access point and modeling the channel, each access point calculates the signal covariance matrix based on a probabilistic algorithm. S1.

2. Define the product of the activity state and the transmit power as an activity parameter. Set a threshold. When the activity parameter is greater than the threshold, the state is considered active. When it is less than the threshold, the state is considered inactive. S1.

3. Use the covariance matrix to obtain the maximum likelihood estimation function for the activity parameter. Estimate the active user status by maximizing the maximum likelihood estimation function, and find the activity parameter that makes user activity most likely to occur in the network. S1.

4. Establish the negative logarithm of the minimum maximum likelihood estimation function for subsequent minimization optimization to ensure the maximization of probability while simplifying the optimization process.

5. The method for detecting active users based on access point clustering according to claim 4, characterized in that: The S2 specifically includes the following contents: S2.

1. Based on the K-means clustering algorithm, after specifying the number of clusters, the K-means internal iteration is used to find the optimal cluster center under the current number of access points, and the geographical location is used to plan the access point clusters; S2.

2. After clustering is completed, calculate the weighted average large-scale fading coefficient of all access points in the cluster; S2.

3. Select the cluster with the largest weighted average large-scale fading coefficient for the user, establish a connection with it, and perform active user detection modeling according to the covariance algorithm. Replace the original large-scale fading coefficient with the weighted average large-scale fading coefficient and minimize the negative logarithm of the maximum likelihood estimation function to achieve active user detection.

6. The method for detecting active users based on access point clustering according to claim 5, characterized in that: The S3 specifically includes the following contents: The negative logarithm of the maximum likelihood function containing the covariance matrix is constructed as the cost function in the coordinate descent method. In each iteration, an active parameter is selected for optimization, and the other parameters are kept unchanged. The active parameter optimization process includes: Calculate the derivative of the active parameter and set its derivative to 0 to find the optimal value that minimizes the cost function; All active parameters are gradually adjusted and updated until all active parameters converge and the objective function no longer changes significantly. Finally, the active parameters that maximize the maximum likelihood function are obtained, which is defined as the optimal solution.

7. A computer device, characterized in that: The computer device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, and the instruction, program, code set, or instruction set is loaded and executed by the processor to implement the active user detection method based on access point clustering as described in any one of claims 2-6.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores at least one instruction, at least one program, code set, or instruction set, and the instruction, program, code set, or instruction set is loaded and executed by the processor to implement the active user detection method based on access point clustering as described in any one of claims 2-6.

Images