Gaussian Process Modeling of UWB Multipath Components

In the field of radio channel modelling, the consideration of ultra-wideband (UWB) signals allows for a channel description that can be decomposed into deterministic multipath components (MPCs) and dense / diffuse multipath. In this paper, deterministic MPCs are studied. A Gaussian Process (GP) model is used as a regression tool to analyze and model the MPC amplitude behavior, based on channel measurements. In this regard, the dependency of the amplitude on the incident angle of the signal at a reflecting surface is analyzed. The channel model is related to kernel functions used in GP that describe the similarities between different data points. Based on the model parameters, the amplitudes of MPCs are predicted at other positions of the receiver that have not been measured in the experiment. The method can then be used to predict a UWB channel impulse response at an arbitrary position in the environment.