Analysing Vibrotactually Stimulated EEG Signals to Comprehend Object Shapes

Abstract

Tactile feedback has the capability of reducing the workload on the visual channel, during visual feedback in brain-computer interfaces (BCIs). It is requisite to analyse the brain signals corresponding to the tactile stimulations. This work is aimed at analysing the brain signals while the users are vibrotactually stimulated. The brain signals are acquired non-invasively by electroencephalography (EEG), while brushless coin-type vibration motors are actuated in particular patterns to convey the object shape information on subjects’ skin surface in form of vibrations. The acquired EEG signals are pre-processed to eliminate the effect of various types of noises and to extract the EEG signals corresponding to relevant frequency bands. Adaptive autoregressive (AAR) parameters are extracted from the pre-processed EEG signals and are finally classified by Naive Bayesian $(NB)$ approach, in order to recognize the vibratotactually stimulated object shapes from brain signals. In addition to the classifier output, subjects’ verbal responses about the object shape they perceived are also noted for validation. Three successive sessions of shape recognition from vibrotactile pattern show an improvement in EEG classification accuracy from 63.75% to 74.37%, and also depicted learning of the stimulus from subjects’ psychological response which is observed to increase from 75% to 95%. This observation substantiates the learning of vibrotactile stimulation in user over the sessions which in turn increases the system efficacy.