Finite mixture models for ultrawideband channels


In this paper we estimate finite mixture models (FMM) to describe the statistics of the ultrawideband (UWB) channel amplitudes. Various combinations of Rayleigh, Nakagami, Weibull, and Lognormal distributions are used to form the constituent probability density functions (PDFs) of the FMMs. The FMMs are identified using the Stochastic Expectation Maximization (SEM) algorithm. Akaike's Information Criterion is used to compare the quality of data fit provided by the FMMs and models containing only one distribution (non-mixtures). The results indicate that UWB channel amplitude statistics are best represented by mixtures of Rayleigh, Lognormal and Weibull PDFs.

Publication Title

Proceedings of SPIE - The International Society for Optical Engineering