Identification of Affective Mental Activity Based on Multichannel EEG Signals

Efficient and effective human-robot collaboration depends on the capacity of both sides to ensure robust communication and mutual awareness of actions, intentions, context, timing, and numerous extra factors that precondition specific behaviors. On the machine side, this would require an intelligent Human-Machine Interface (iHMI), which among other functionalities is also sensitive to the human cognitive and affective states. In this regard, we propose a method for the automated identification of affective mental activity based on EEG signals. The proposed method is computationally lightweight as it operates directly on the samples of the time-domain signal, without the need for complex preprocessing, frequency-domain transformation, or another feature extraction steps. The method is validated in an experimental setup based on the DEAP dataset and was shown to significantly outperform two other setups. The experimental results support the usefulness of the proposed method and we deem it will facilitate the development of iHMI, which are sensitive to the four major categories of affective states as defined in the arousal-valence space.