Contraction, Smoothness, and Low-Pass Filtering

Abstract

We introduce a generalized definition for "low-pass" filters that covers time-varying and nonlinear systems under the same umbrella. We show that the qualitative concept of signal smoothing can be made precise through the concept of contractions in probabilistic metric spaces. For illustration, we consider classical linear time-invariant low-pass filters, nonlinear median filters, and time-varying guaranteed maximum delay schedulers employed in communication systems.