On the reliability of multipath cluster estimation in realistic channel data sets
Christian Schneider, Maysam Ibraheam, Stephan Häfner, Martin Käske, Matthias Hein, Reiner S. Thomä · 2014
For the parametrization of geometry based stochastic channel models as from the IST-WINNER or COST 273/IC1004 initiatives large data analysis from channel sounding campaigns play an important role. Whereby the reliability of cluster charaterisation as a post-processing step after the high resolution multipath estimation exhibits a crucial issue. In this contribution a framework for evaluation and development of different cluster algorithms is discussed. Furthermore a novel hierarchical algorithm is introduced and compared to standard K-means and Fuzzy-C-means algorithms. Whereby the new algorithm outperforms the standard algorithms wrt. increasing number and size of clusters.