Gaussian Mixture Based Context-Aware Short-Term Characterization of Wireless Channels

Abstract

5G wireless communication technologies aim at simultaneously achieving energy efficiency and spectral efficiency. 5G also demands high communication reliability. In this context, fine-grained temporal characterization of wireless channel can be used to enhance both. To this end, we propose a novel context-aware characterization of the temporally-varying wireless channel. Our characterization of temporal variation of the channel is based on the method of finite mixture of Gaussian distributions. However, unlike the classical Gaussian mixture model, the proposed characterization does not use an iterative algorithm for its parameter estimation; it depends on the current channel state and its statistics. Based on this characterization we estimate the quantity of data that can be transferred over the channel in a time interval without knowing the actual channel state in that duration. We propose an application context dependent upper bound on the time interval over which this estimation can be made. Our numerical results demonstrate that the present channel state plays a crucial role. When the proposed characterization is used in the context of channel adaptive communication, energy efficiency obtained is as high as 3.15 times over its nearest approach. A nontrivial trade-off between energy efficiency and precision of the proposed characterization is also investigated.