A Clustering Algorithm Based on Joint Kernel Density for Millimeter Wave Radio Channels

Cluster-based channel modeling has gradually become a trend, since it can balance modeling accuracy and complexity. In this paper, we propose a density-based clustering algorithm to cluster channel multi-path components (MPCs), which considers the statistical characteristics of the parameters when calculating the density with joint kernel equation. To validate the performance of the algorithm, both simulation and a millimeter-wave based urban-microcell channel measurement are performed. Compared with KPowerMeans, the results of simulation show that the proposed algorithm can identify clusters with a higher success rate validated by Fowlkes-Mallows score (FMI), and measurement-based clustering results have better intra-cluster compactness and inter-cluster separation validated by Calinski-Harabasz (CH) index and Davies-Bouldin (DB) criterion. Furthermore, the proposed algorithm does not require preset the number of clusters, which makes it more intelligent.