A Novel Clustering Method Based on Density Peaks and Its Validation by Channel Measurement

Hanqing Xue, Lei Tian , Yuxiang Zhang, Chao Wang, Jianhua Zhang , Wei Li, Hewen Wei
2018 European Conference on Antennas and Propagation
Publish Year:

Abstract: In this paper, cluster centers are defined as local maxima in the density which is called density peaks. We use an algorithm based on density peaks to cluster multipath components (MPCs) whose parameters are estimated by the space-alternating generalized expectation-maximization (SAGE) algorithm.To validate its performance, we apply it to the data from a channel measurement in an indoor scenario at 28 GHz and compare it to the widely used KPowerMeans algorithm with the same distance metric of multiple component distance (MCD). Through the results, we find it can cluster MPCs more accurately validated by Calinski-Harabasz (CH) and Davies-Bouldin (DB) indices. More importantly, it overcomes the shortcoming of setting the number of clusters manually because we can determine the cluster centers automatically. Besides, the efficiency of DP clustering is far superior to KPowerMeans because what it needs is only a distance matrix and we only need to compute it once. At the end, statistical analysis of clusters are presented to provide insights in channel modeling of millimeter wave.


pdf writencite-old

Leave a Reply

Your email address will not be published. Required fields are marked *

7 − = five

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>