Investigating mechanical properties of a fabric-based affective haptic display through electrodermal activity analysis

The aim of this study is the development of a classification system able to discern between two levels of velocity of a caress-like haptic stimulus, through information gathered from the analysis of the Electrodermal Activity (EDA) dynamics. We designed and performed an experiment where EDA signals were acquired during caress-like stimuli conveyed to 32 healthy volunteers (16 females) by means of four different fabrics (hemp, burlap, velvet and silk) and at two velocity levels (9.4 mm/s and 65 mm/s). CvxEDA model was used to process the EDA signal and extract features from both tonic and phasic components. The feature set was used as an input to a K-NN classifier implementing a leave-one-subject-out procedure. Considering all fabrics, results show an accuracy of velocity recognition between 91.07% and 96.43%. Conversely, poor accuracy was achieved considering the fabric classification. Results also suggest that caress velocity significantly affects EDA dynamics regardless of the typology of fabrics. This is probably due to the fact that velocity is strictly related to the valence of the affective stimuli.