Principal component analysis of galvanic skin responses

The galvanic skin response (GSR) is a simple method of capturing the autonomic nerve response as a parameter of the sweat gland function. Any stimulus capable of an arousal effect can evoke the response and the amplitude of the response is more dependent on the surprise effect of the stimulus than on the physical stimulus strength. In this paper principal component analysis (PCA) is used for the analysis of the evoked GSRs. Basis functions are obtained from the eigendecomposition of the data correlation matrix. Because PCA is the best mean square fit of a set of orthogonal functions to the set of measurements, the solution will depend upon the nature of measurements. The dimensionality of measurements can be estimated by the number of basis functions needed to estimate measurements in a certain accuracy. Hence the eigenvalues, corresponding to used basis functions, are a measure of similarity. The method was tested using 20 healthy subjects and 13 psychotic patients. 11 surprising auditory stimuli were delivered at irregular intervals and evoked GSRs were recorded from the hand. Observed similarities between adjacent waveforms were more remarkable within healthy subjects.