Cepstral distance measures of hormone concentration time series

Abstract

In this paper we present a class of time series distance measures based on the difference of their cepstral transformations. We emphasise the convenience of the proposed distance measure in the cases when the time series can be treated as output of a linear system driven with a quasi-periodic stochastic signals. In order to illustrate the cepstral time series distance measure we applied them in cluster and multidimensional scaling analysis of daily hormonal secretion fluctuation series taken from a group of patients before and after surgery. © 1998 IMIA. All rights reserved.