Using entropy to build efficient FIR digital filters

In this paper, we introduce the notion of entropy in the context of digital filter sensitivity. In particular, we discuss its application in several FIR filter structures. We compare the performance of our introduced entropy sensitivity to the well-known power sensitivity measure commonly used in this kind of work. We show that while the entropy-predicted performance nearly conforms to current expected practice, some interesting issues related to filter order occur. Furthermore, the entropy measure, based as it is in time-frequency distribution (concentration) measures, more accurately predicts behavior as the sampling rate increases, or as the resolution of the filter specification increases.