Metrological Evaluation of the Compressive Sampling Approach Applied to ElectroDermal Activity Signals in response to Audio Stimuli

The ElectroDermal Activity (EDA) signal, recorded as the electrical conductance between a pair of electrodes placed over a person’s skin, and linked to a person’s emotional arousal, consists of a tonic component superposed by multiple Skin Conductance Responses (SCRs). Each SCR, in turn, results from a corresponding SCR event which is not directly observable from the measured EDA signal although its informative content is very useful in a wide range of applications. Thus, several approaches have been proposed to extract the SCR events from the EDA signal, but the problem has not been solved yet. In this work, the effectiveness of a Compressed sensing (CS)-based approach for the extraction of SCR events from EDA signals measured with a wrist-worn wearable device, has been investigated. Once the sparse SCR events signals have been extracted from real EDA signals acquired during a sound stimuli experiment, the SCR onsets detection accuracy of this CS-based approach has been assessed against more traditionally used methods. The attained results show that, overall, there are no relevant differences but a small overestimate in the number of detected SCR onsets with the proposed approach compared to the other strategies, demonstrating the goodness of this CS-based method.