Clustering in 3D MIMO Channel: Measurement-Based Results and Improvements

In this paper, we perform 3-Dimensional (3D) clustering based on the Outdoor-to-Indoor (O2I) wideband 3D multiple-input-multiple-output (MIMO) channel measurement at 3.5 GHz. Clusters are identified by KPowerMeans algorithm. Based on analysis on clustering results, we modified the definition of Multiple component distance (MCD) to split the bounding of azimuth and elevation, which can obtain larger number of clusters and the clusters are more intra- compact and inter-separated. Then, Calinski-Harabasz (CH) and Davies-Bouldin (DB) indices are used to further validate the proposed MCD. Finally, intra cluster and inter cluster statistics are both provided, which provides insights in 3D MIMO channel modeling.