Wireless channel multipath clustering method based on spatial transformation and dimension weighting

Abstract

The invention discloses a wireless channel multipath clustering method based on spatial transformation and dimension weighting, which can effectively realize clustering of multipath components in a wireless channel, and specifically comprises the following steps: 1) setting the number of multipath clusters and the initial centroid position; 2) carrying out initial iteration, calculating the Euclidean distance between a sample point and the centroid of each cluster to obtain a membership matrix, a spatial transformation matrix and a dimension weighting coefficient, and updating the position of the centroid; 3) for subsequent iteration, firstly performing spatial transformation on the multipath data set according to a spatial transformation matrix obtained by the previous iteration, then calculating a weighted Euclidean distance between a sample point and each cluster centroid, and finally updating a membership matrix; and 4) carrying out loop iteration until convergence, and carrying out multipath cluster division by using the finally obtained membership matrix. Compared with a traditional division type clustering method, the method provides a clustering result which is more accurate, and can be used for supporting wireless channel modeling work based on clusters.

Description

Technical Field

[0001] The present invention belongs to the field of wireless channel research, and particularly relates to a wireless channel multipath clustering method based on space transformation and dimension weighting. Background Art

[0002] In wireless communication, when a signal encounters obstacles during propagation, it will experience phenomena such as reflection, diffraction, and scattering, and thus reach the receiving end through different paths. The signal transmitted through each path is called a multipath component (MPC). The core of wireless channel modeling lies in accurately describing the multipath components in the channel. A large number of channel measurement results show that multipath components usually exist in the form of clusters, and these clusters have similar characteristics in dimensions such as power, delay, and angle. Therefore, the cluster-based channel model has become an important direction in wireless channel modeling research. Clustering of multipath components is the first step in channel modeling. In the early stage, researchers judged the distribution of clusters through visual inspection, which was feasible when the number of multipath components was small. However, with the expansion of the scale of channel measurement, the limitations of visual inspection gradually emerged, because the human eye cannot process high-dimensional data and there is subjectivity in the division of clusters, thus affecting the accuracy.

[0003] In response to the above challenges, some scholars have applied clustering algorithms in machine learning to the clustering of MPCs in wireless channels. The fuzzy C-means (FCM) algorithm is a soft partitioning clustering method. The advantages of this method are fast convergence speed and highly reproducible results. However, existing partitioning clustering methods cannot effectively identify non-spherical clusters, especially when the sizes of clusters are different. To fill this gap, the present invention proposes an improved fuzzy clustering method. Summary of the Invention

[0004] Object of the Invention. The object of the present invention is to provide a wireless channel multipath clustering method based on space transformation and dimension weighting to solve the problem that existing partitioning clustering methods are difficult to accurately identify non-spherical clusters of different sizes.

[0005] Technical Solution. To solve the above technical problems, the present invention proposes a wireless channel multipath clustering method based on space transformation and dimension weighting, including the following steps:

[0006] Step 1: Set the number K of clusters of wireless channel multipath components;

[0007] Step 2: Initialize the centroid positions C = [c k K×1 ;

[0008] Step 3: Calculate the Euclidean distance d between the nth multipath and the centroid of the kth cluster k,n ​, and obtain the membership matrix \(U = [u k,n K×N ;

[0009]

[0010] Step 4: Calculate the spatial transformation matrix \(S\) and the weighted coefficient \(W = [w k,i K×d , where \(d\) represents the dimension of the sample space, and update the cluster centroid position \(C\);

[0011] Step 5: For all clusters, perform spatial transformation and dimensional weighting on the original multipath data set in sequence, and update the membership matrix \(U\) based on the transformed multipath data set;

[0012] Step 6: Determine whether the convergence condition is reached. If it has converged, obtain the final clustering result; otherwise, return to Step 4 and continue to update and iterate \(C\), \(S\), and \(W\) according to the membership matrix \(U\).

[0013] Furthermore, Step 4 specifically includes the following steps:

[0014] Step 4.1: Calculate the covariance matrix of each cluster. The calculation method of the covariance matrix is:

[0015]

[0016] where \(x n represents the \(n\)th multipath component, \(c k represents the centroid of the \(k\)th cluster, \(u k,n is the membership of the \(n\)th multipath component to the \(k\)th cluster, \(N\) represents the total number of multipaths, and \(m\) is the fuzzy coefficient;

[0017] Step 4.2: Calculate the eigenvectors of the within-cluster covariance matrix;

[0018] Step 4.3: Calculate the spatial transformation matrix required for each cluster according to the eigenvectors. The calculation method of the spatial transformation matrix is:

[0019] S k =[v 1 ,…,v d -1

[0020] where \(v i is the eigenvector of the within-cluster covariance matrix, and \(d\) represents the dimension of the sample space;

[0021] Step 4.4: Calculate the dimensional weighting coefficient according to the membership matrix. The calculation method of the dimensional weighting coefficient is:

[0022] ​​​

[0023] where d k,n,i is the component of the distance from x n to the k-th centroid in the i-th dimension, w k,i is the weight of this dimension, γ is a proportionality parameter, and p serves a similar role to m and is also a fuzziness coefficient;

[0024] Step 4.5: Update the centroid position of the cluster according to the membership matrix. The centroid update equation is:

[0025]

[0026] Furthermore, Step 5 specifically includes the following steps:

[0027] Step 5.1: For the k-th cluster, according to the obtained spatial transformation matrix S k , perform spatial transformation on the original multipath data set X and the centroid c k of this cluster. The calculation method is:

[0028] X T = S k ·X

[0029] c k T = S k ·c k

[0030] Step 5.2: Calculate the weighted Euclidean distance between the transformed multipath data set X T and c k T . The calculation method is:

[0031]

[0032] Step 5.3: Update the membership matrix according to the weighted Euclidean distance. The calculation method is:

[0033]

[0034] Step 5.4: Successively perform Steps 5.1 to 5.3 for all clusters to obtain the updated membership matrix.

[0035] Furthermore, Step 6 specifically includes the following steps:

[0036] Step 6.1: If the convergence condition is not reached, continue to update the iteration parameters according to Step 4. The convergence condition is:

[0037] ∑||C (i+1) - C (i) || 2 < ε

[0038] Among them, C (i) is the centroid set at the i-th iteration, and ε is the preset iteration error;

[0039] Step 6.2: If the convergence condition has been reached, then according to the membership matrix, each multipath component is divided into the cluster with the highest membership degree.

[0040] In addition, the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of a wireless channel multipath clustering method based on space transformation and dimension weighting are implemented.

[0041] In addition, a computer terminal device includes a processor, a memory, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of a wireless channel multipath clustering method based on space transformation and dimension weighting are implemented.

[0042] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0043] Based on the existing fuzzy clustering method, the present invention proposes an FCM algorithm based on space transformation and weighting for wireless channel multipath clustering. First, when the FCM algorithm is used for clustering by applying the Euclidean distance, it is significantly affected by the dimensional variance of the data set and the correlation between dimensions. To overcome this problem, the process of calculating the Mahalanobis distance is divided into two steps: space transformation and dimension weighting. This method solves the difficulty of directly applying the Mahalanobis distance in FCM clustering and overcomes the defect that the GK algorithm tends to divide clusters of the same size when indirectly applying the Mahalanobis distance metric. In summary, the present invention is an effective multipath clustering method, providing a prerequisite for further realizing accurate channel modeling. Brief Description of the Drawings

[0044] Figure 1 is the model block diagram of the wireless channel multipath clustering method based on space transformation and dimension weighting provided in Embodiment 1 of the present invention;

[0045] Figure 2 is the schematic diagram of the principle of space transformation and dimension weighting in Embodiment 1 of the present invention;

[0046] Figure 3 is the schematic diagram of the original multipath components in Embodiment 1 of the present invention;

[0047] Figure 4 is the schematic diagram of the multipath components after clustering in Embodiment 1 of the present invention. Detailed Embodiment

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0049] Embodiment 1:

[0050] Refer to Figure 1 , this embodiment provides a wireless channel multipath clustering method based on space transformation and dimension weighting, including the following steps:

[0051] Step 1, set the number K of multipath component clusters of the wireless channel;

[0052] Step 2, initialize the centroid positions C = [c k K×1 ;

[0053] Step 3, calculate the Euclidean distance d k,n between the nth multipath and the centroid of the kth cluster, and obtain the membership matrix U = [u k,n K×N according to the following formula;

[0054]

[0055] Step 4, calculate the space transformation matrix S and the weighting coefficients W = [w k,i K×d according to the membership matrix U, where d represents the dimension of the sample space, and update the centroid position C;

[0056] Step 5, for all clusters, sequentially perform space transformation and dimension weighting on the original multipath data set, and update the membership matrix U based on the transformed multipath data set;

[0057] Step 6, determine whether the convergence condition is reached. If it has converged, obtain the final clustering result; otherwise, return to Step 4 and continue to update and iterate C, S, and W according to the membership matrix U.

[0058] Further, Step 4 specifically includes the following steps:

[0059] Step 4.1, calculate the covariance matrix of each cluster, and the calculation method of the covariance matrix is:

[0060]

[0061] where x n represents the nth multipath component, c​​​k represents the centroid of the k-th cluster, u k,n is the membership degree of the n-th multipath component to the k-th cluster, N represents the total number of multipaths, and m is the fuzzy coefficient;

[0062] Step 4.2: Calculate the eigenvectors of the within-cluster covariance matrix;

[0063] Step 4.3: Calculate the spatial transformation matrix required for each cluster according to the eigenvectors. The calculation method of the spatial transformation matrix is:

[0064] S k =[v 1 ,…,v d -1

[0065] where v i is the eigenvector of the within-cluster covariance matrix, and d represents the dimension of the sample space;

[0066] Step 4.4: Calculate the dimension weighting coefficient according to the membership matrix. The calculation method of the dimension weighting coefficient is:

[0067]

[0068] where d k,n,i is the component of the distance from x n to the k-th centroid in the i-th dimension, w k,i is the weight of this dimension, γ is a proportionality parameter, and the role of p is similar to that of m, and it is also a fuzzy coefficient;

[0069] Step 4.5: Update the centroid position of the cluster according to the membership matrix. The centroid update equation is:

[0070]

[0071] Furthermore, Step 5 specifically includes the following steps:

[0072] Step 5.1: For the k-th cluster, according to the obtained spatial transformation matrix S k , perform spatial transformation on the original multipath data set X and the centroid c k of this cluster. The calculation method is:

[0073] X T =S k ·X

[0074] c k T =S k ·c k

[0075] Step 5.2: Calculate the transformed multipath data set X​T The weighted Euclidean distance between it and c k T is calculated as follows:

[0076]

[0077] Step 5.3: Update the membership matrix according to the weighted Euclidean distance, and the calculation method is:

[0078]

[0079] Step 5.4: Perform Steps 5.1 to 5.3 on all clusters in sequence to obtain the updated membership matrix.

[0080] Furthermore, Step 6 specifically includes the following steps:

[0081] Step 6.1: If the convergence condition is not reached, continue to update the iteration parameters according to Step 4. The convergence condition is:

[0082] ∑||C (i+1) -C (i) || 2 <ε

[0083] where C (i) is the centroid set at the i-th iteration, and ε is the preset iteration error;

[0084] Step 6.2: If the convergence condition has been reached, divide each multipath component into the cluster with the highest membership according to the membership matrix.

[0085] In addition, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a method for wireless channel multipath clustering based on space transformation and dimension weighting are implemented.

[0086] In addition, a computer terminal device includes a processor, a memory, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of a method for wireless channel multipath clustering based on space transformation and dimension weighting are implemented.

[0087] In summary, the method for wireless channel multipath clustering based on space transformation and dimension weighting established by the present invention clusters the multipath components of the wireless channel by using the Mahalanobis distance metric, has a more relaxed assumption about the cluster shape, can better adapt to irregularly shaped clusters, and finally realizes the accurate division of multipath clusters, providing an important prerequisite for wireless channel modeling.

[0088] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field according to the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art should fall within the protection scope determined by the claims.

Claims

1. A wireless channel multipath clustering method based on spatial transformation and dimensional weighting, characterized in that: The following steps are involved: Step 1, setting the number K of multipath component clusters of the wireless channel; Step 2: Initialize the centroid positions C of K multipath clusters = [c k ] K×1 ; Step 3: Calculate the Euclidean distance d between the nth multipath and the kth cluster centroid k,n , and the membership matrix U = [u k,n ] K×N ; Step 4: Calculate the spatial transformation matrix S and weighting coefficient W = [w k,i ] K×d , where d represents the dimension of the sample space, and updates the cluster centroid position C; Step 5: For all clusters, spatial transformation and dimension weighting are performed on the original multipath data set in turn, and the membership matrix U is updated based on the transformed multipath data set; Step 6: Determine whether the convergence condition is met. If converged, the final clustering result is obtained. Otherwise, return to step 4 and continue to update C, S, and W according to the membership matrix U.

2. A wireless channel multipath clustering method based on spatial transformation and dimensional weighting according to claim 1, characterized in that: Step 4 specifically includes the following steps: Step 4.1: Calculate the covariance matrix of each cluster. The covariance matrix is ​​calculated as follows: Among them, x n represents the nth multipath component, c k represents the centroid of the kth cluster, u k,n is the membership degree of the nth multipath component to the kth cluster, N represents the total number of multipaths, and m is the fuzzy coefficient; Step 4.2, calculate the eigenvector of the intra-cluster covariance matrix; Step 4.3: Calculate the spatial transformation matrix required for each cluster based on the eigenvector. The spatial transformation matrix is ​​calculated as follows: S k =[v1,…,v d ] -1 Among them, v i is the eigenvector of the intra-cluster covariance matrix, and d represents the dimension of the sample space; Step 4.4: Calculate the dimension weight coefficient according to the membership matrix. The dimension weight coefficient is calculated as follows: Among them, d k,n,i For x n The component of the distance to the kth centroid in the i-th dimension, w k,i is the weight of the dimension, γ is a proportional parameter, and p has a similar function to m, which is also a fuzzy coefficient; Step 4.5: Update the centroid position of the cluster according to the membership matrix. The centroid update equation is:

3. The wireless channel multipath clustering method based on spatial transformation and dimensional weighting according to claim 1, characterized in that: Step 5 specifically includes the following steps: Step 5.1: For the kth cluster, according to the obtained spatial transformation matrix S k , for the original multipath data set X and the centroid c of the cluster k Perform spatial transformation, the calculation method is: X T =S k ·X c k T =S k ·c k Step 5.2: Calculate the transformed multipath data set X T With c k T The weighted Euclidean distance between is calculated as: Step 5.3: Update the membership matrix according to the weighted Euclidean distance. The calculation method is: Step 5.4: Perform steps 5.1 to 5.3 for all clusters in turn to obtain the updated membership matrix.

4. The wireless channel multipath clustering method based on spatial transformation and dimensional weighting according to claim 1, characterized in that: Step 6 specifically includes the following steps: Step 6.1: If the convergence condition is not met, continue to update the iterative parameters as described in step 4. The convergence condition is: ∑||C (i+1) -C (i) || 2 <e Among them, C (i) is the centroid set at the i-th iteration, and ε is the preset iteration error; Step 6.2: If the convergence condition has been reached, each multipath component is divided into the cluster with the highest membership according to the membership matrix.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a wireless channel multipath clustering method based on spatial transformation and dimensional weighting described in claims 1-4 are implemented.

6. A computer terminal device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of a wireless channel multipath clustering method based on spatial transformation and dimensional weighting according to claims 1 to 4 are implemented.

Images