Recognition of affective haptic stimuli conveyed by different fabrics sing EEG-based sparse SVM

This study reports on the modification of electroencephalographic (EEG) power spectra as a response to quantifiable affective haptic stimuli. Ten healthy subjects (5 females) were enrolled to interact with a haptic device able to mimic caress-like stimuli by using different fabric strips at different velocities. More specifically, 2 velocity levels and 4 different fabrics, i.e. hemp, burlap, velvet and silk, were combined to deliver different affective elicitations. Subjects were also asked to evaluate the haptic stimuli in terms of arousal (activation/deactivation) and valence (pleasantness/unpleasantness) scores. Results are reported in terms of p-value topographic maps, revealing that only silk (pleasant fabric) and burlap (unpleasant fabric) were able to induce a suppression of EEG rhythms in all the frequency bands. Moreover, silk strip was able to induce a significant activation in both lateral and contralateral areas at lower velocity (valence) stimulation. In addition, a pattern recognition approach was proposed to automatically recognize pleasant and unpleasant elicitation using a L1-regularized SVM, showing an average accuracy of 70%.