Multiresolution Karhunen Loéve analysis of galvanic skin response for psycho-physiological studies

Abstract

This paper proposes the use of Multiresolution Karhunen Loeve (MR-KL) analysis to analyse Galvanic Skin Response (GSR) Signals. GSR signal, which is considered the golden standard in peripheral neuro-physiological and psycho-physiological studies, can be represented by two functions representing either the large and small scale variabilities of the series. The large scale variability, related to the tonic component, is modelled by smoothing splines while the small scale variability, related to the phasic component, is modelled by a MR-KL expansion. The recognition of these two components in a real data set is mandatory in neuro-psycho-physiology and we demonstrate the efficacy of the proposed methodology by analysing a pilot experiment on a group of 13 subjects.