2-D Predictive Filters for Polynomial Signals With Applications to Wind Profiler Data

Abstract

Polynomial predictors are known for their ability, in the absence of noise, to exactly predict a future value of a polynomial signal of a fixed order. One-dimensional filtering is a mature field and sophisticated filter design methods have already been heavily studied. Real world 2-D and higher order datasets are widely available for a multitude of applications. Thus, it is interesting to extend the existing one-dimensional polynomial predictors, e.g. Heinonen-Neuvo filter, to higher dimensional spaces. In this paper, we propose a novel 2-D polynomial predictor and evaluate its performance on a newly generated wind speed dataset.