Using entropy for dimension reduction of tactile data

Tactile sensing arrays for robotic applications become more and more popular these days. This allows us to equip robots with sensing abilities similar to those of our human skin. This paper presents an approach to tactile-based recognition of objects and evaluates the utility of various feature extractors for tactile processing. Extracting these features from a tactile database, we describe a system that combines a discretization step with the well-known C4.5 algorithm in an object classification task. We analyse the usefulness of the features in terms of entropy-based considerations taking into account the generated decision trees and report our results that give important hints for feature selection.