Surface roughness classification in dynamic touch using EEG signals

This study uses deep learning and EEG signals to classify surface roughness, an active yet less explored field of research that employs a robotic device to conduct tactile trials. This automated system utilizes a rotating wheel to position items with varying roughness levels, including smooth, semi-rough, and rough. The experiment was arranged such that subjects could use their index fingers on both hands to touch items embedded in the wheel. Each item makes contact with the index finger in both static and dynamic touch modes, and participants’ brain data (EEG) are collected during the experiment. The recorded EEG is then utilized to create a model capable of detecting the roughness level of surfaces. The proposed TactileNet model, which is based on convolutional neural networks, classified the roughness level of touched surfaces with an accuracy of 95.67 percent. According to the comparative analysis, the proposed approach outperforms state-of-the-art results in terms of accuracy. Additionally, unlike previous research, the suggested method does not require manual feature extraction from EEG data.