Optimization of Signal Processing Parameters in Psychophysiological Studies on the Example of GSR and PPG

When analyzing physiological signals, the problem of setting data processing parameters arises due to the blurring of the boundary between signal and noise properties, as well as the fundamental lack of objective criteria for the quality of data processing in psychophysiology. This paper describes an approach to optimizing processing parameters on the example of galvanic skin response (GSR) and photoplethysmogram (PPG), based on the use of stimuli that are significant for a person, selected on the basis of biographical data, which can be considered as criteria validation. As a metric for the optimization, we used the frequency of coincidence of the stimuli identified as a result of the analysis with the a priori given ones (human names, including the name of the volunteer, and also visit cards selected by the volunteer). GSR and PPG signals were recorded using an MRI-compatible polygraph under conditions of functional magnetic resonance imaging (N=46 volunteers). In the first part of the work, optimization of frequency filters and analysis intervals (epochs) was performed. It has been established that the following processing parameters are optimal for analyzing the amplitude properties of the GSR signal: first-order Butterworth filters, frequency range is 0.025-0.25 Hz, interval of analysisis1-7 s from a stimulus. To analyze the PPG signal using the length of the curve, the following processing parameters are optimal: second-order Butterworth filters, frequency range is 1.25—12.5 Hz, interval of analysis is 3—10 s from a stimulus. Using the same criterion, several alternative signal processing methods were tested: change in the amplitude of the GSR signal over the analysis interval compared to the classical method by the amplitude maximum relative to the baseline; several types of ranking of reactions within a block of stimuli compared to simple averaging of all responses. The parameters and methods of processing of the GSR and PPG signals obtained in the work demonstrate universality in relation to the variety of initial data and could be applicable in applied and fundamental research. The general approach described in the work can also be used to optimize the processing parameters of other physiological signals including fMRI.