A modified domain deformation theory on 1-D signal classification

Abstract

In this paper, classification of one-dimensional (1-D) signals is accomplished using domain deformation theory. We use a metric defined on a domain deformation for measuring the distance between signals. By introducing a newly defined metric space, the assumption that domain deformation is applicable only to continuous signals is removed such that any kind of integrable signal can be classified. This method also has an advantage over the L2 metric, because the similarity of one-dimensional signals can be better measured for the purpose of classification.