EEG-Based Texture Classification During Active Touch

Abstract

Modern approaches to providing haptic feedback focus mainly on robotic manipulators, vibrators, and tactors. This type of feedback tends to be cumbersome and limited to a small number of contact points. On the contrary, electrotactile displays are compact and wearable, and recent discoveries demonstrate that naturalistic sensations of touch can be provided by electrical stimulation of peripheral nerves. Haptic feedback is essential for daily activities, as it allows us to become aware of our surroundings. The aim of this study is to develop techniques to extract EEG features that are markers for real world haptic interactions. The extracted EEG features will be used to model the EEG evidence that will later be employed in the closed-loop guidance system to adaptively control the electrical stimulation. In this work, twelve healthy subjects were recruited to perform a tactile stimulation experiment. Each subject was instructed to use their index finger to rub or tap 3 textured surfaces having varying levels of roughness (smooth flat, medium rough, and rough). EEG and force data were collected synchronously during each movement condition. Analysis of the EEG data showed that the amplitude of the EEG segment and the total power in the Mu (815 Hz) and Beta (1630 Hz) bands could be used to identify the roughness of textures using EEG data. We used a 10-fold cross validation to train a 3-class Support Vector machine classifier with chance level of 33%. The results show that it is possible to discriminate EEG activity with very high accuracy among surfaces with different textures.