Abstract—Wireless multipath propagation causes different paths taken by the signal due to interacting objects present in the environment producing multipath components (MPCs). Cluster-based channel models characterize the wireless channel, and different approaches are utilized to cluster the MPCs. Data mining requires different techniques such as visualization to extract important information and find patterns and clusters in the data. A Graphical User Interface (GUI) is developed in this paper to aid the visualization and the manual clustering of MPCs using t-distributed Stochastic Neighborhood Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP). The clustering results of Simultaneous Clustering and Model Selection (SCAMS) are used in this paper. The datasets are embedded into low-dimensional projection and are manually re-clustered. The manual clustering was performed visually and interactively, which achieves a higher Jaccard membership index with a low value of 0.3368, a median of 0.4697, and a high value of 0.8884 for all the datasets.
 

Index Terms—Graphical user interface, multipath clustering, interactive clustering, wireless propagation, dimensionality reduction

Cite: Emmanuel T. Trinidad and Lawrence Materum, "High-Accuracy Visualization-Based Grouping of MIMO Multipath Waves," Journal of Communications vol. 18, no. 1, pp. 68-75, January 2023. Doi: 10.12720/jcm.18.1.68-75