Statistical technique to identify clusters from multi-dimensional measurement data

H. Xiao*, A. G. Burr, L. Hentilä, P. Kyösti

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution


This paper presents a statistical technique to be applied to multi-dimensional (MD) channel estimation data to objectively establish the presence of clusters of the multi-path components (MPCs). The key insight is that by dividing the MD measurement data space into MD analysis regions (MDARs), we can compare the distributions both of the number of the MPCs and the total power for the MPCs within a MDAR, with those distributions when the MPCs are random and uncorrelated in the MD data space. Hence we may confirm or reject the presence of significant clustering. This statistical technique has been applied to high-resolution indoor multi-input multi-output (MIMO) measurement data at 2.55 GHz to evaluate its applicability, validity and reliability. By analyzing the measurement data using the technique, we found that the statistical technique has good performance in identifying the existence of clusters in two respects, i.e. the probability density functions (PDFs) of both the number and the total power of the MPCs in an MDAR.

Original languageEnglish
Title of host publication2nd European Conference on Antennas and Propagation, EuCAP 2007
Publication statusPublished - 2007
Event2nd European Conference on Antennas and Propagation, EuCAP 2007 - Edinburgh, United Kingdom
Duration: 11 Nov 200716 Nov 2007

Publication series

NameIET Seminar Digest


Conference2nd European Conference on Antennas and Propagation, EuCAP 2007
Country/TerritoryUnited Kingdom


  • Clusters
  • MIMO measurement
  • Poisson distribution
  • Sum of lognormal distribution

Cite this