Error-related potentials in EEG signals: feature-based detection for human-robot interaction

This study explores how to improve the detection of Error-Related Potentials (ErrPs), namely brain signals generated when a person perceives an unexpected action performed by an interacting agent. ErrPs are promising for improving interactions between humans and robots because they offer a way for robots to understand the user's needs and expectations without explicit input. The proposed method aims at characterizing ErrP signals using a wide set of features extracted from electroencephalography (EEG) data, collected from subjects performing different tasks. This feature-based method results more accurate and efficient than traditional approaches, especially when applied to multiple users, or across different experimental setups. This work paves the way to feature-based ErrP detection to enhance human-robot interaction in dynamic environments.