Analysis of galvanic skin responses with principal components and clustering techniques

An advanced method for analyzing the patterning of successive galvanic skin responses (GSR) is presented. The proposed method is based on principal co mponent analysis in which the vector containing the measured signal is pres ented as a weighted sum of orthogonal basis vectors. The method is tested u sing measurements from 20 healthy controls and 13 psychotic patients. For e ach subject, 11 surprising auditory stimuli were delivered to right ear at irregular intervals and evoked GSRs were recorded from the hand. For most o f the healthy controls, there was a clear pattern in successive GSRs, where as within psychotic patients the lack of time-locking of GSRs seemed to be characteristical. These between group differences can be revealed by the pr oposed method. With application to clustering a significant discrimination, with overall correct ratings of 82%, of healthy controls and psychotic pat ients is achieved. A significant fact is that all patients were ranked corr ectly giving the proposed method a sensitivity of 100%.