Understanding human response to tactile stimuli: A Machine Learning approach

Abstract

Whereas understanding human reaction to touch is of great interest in many medical applications, it is still a very unknown field. This research aims to clarify the nature of the relation between endogenous and exogenous attention by analysing electroencephalografic (EEG) data regarding human touch. To this end, data collected from twelve subjects under an experiment based on a variation of the Posner’s cue-target paradigm has been used. After pre-processing, several multi-class classification models based on state-of-the-art machine learning algorithms have been implemented and their accuracy in detecting different experimental conditions have been evaluated. A temporal analysis has also been performed to select the most representative time points. Results showed that although the physical stimuli was identical across conditions, different types of attentional scenarios were classified above chance. Further, the hemisphere contralateral and ipsilateral to the attended side contributed differently, across time, to the accuracy of classification.