Energy based clustering method to estimate channel occupation of LTE in unlicensed spectrum

Abstract

In this article we propose a subtractive histogram clustering method to estimate the number of LTE users based on the uplink energy probability density distribution sensed by an RF front-end. The energy of the signal is estimated and its histogram is analyzed to determinate the number of different distributions. As the energy estimation of the sensed LTE uplink can be modeled by a Gaussian mixture, this allow us to have a priori information that help us to determine the number of users. The lowest value Gaussian distribution can be used to accurately estimate the noise floor and the remaining distributions allow us to estimate the number of LTE users and their received power.