Metrological foundations of emotional valence measurement through an EEG-based system

A contribution regarding the measurability of emotions and a reflection on its main related issues, are provided. The proposed use case is an electroencephalography-based detection system of positive and negative valence states. Some metrological characteristics of the proposed system were considered, i.e. reproducibility, sensitivity, and resolution. The issues concerning the measurability of emotions as lack of experimental, cross-subject, and within-subject reproducibility, as well as uncertainty induced by the adopted stimuli, were highlighted. A theoretical reference model was first identified, namely the circumplex model of affect, which is based on an interval scale. A standardized stimuli set known as Oasis was exploited. Furthermore, an initial screening of the participants was carried out to manage the bias of depressive disorders, and a compatibility analysis was conducted between the experimental sample and the sample exploited by the standardized dataset. The effectiveness of the emotion induction was maximized by choosing a polarized subset of stimuli and an implicit-more controlled mood induction procedure. A Self-Assessment Manikin was employed to verify the effectiveness of the induction procedure. The validity of the proposed method was experimentally proved. EEG signals from 25 healthy subjects were acquired through a 8-channel device. As a result, an average accuracy of 96.1 % in the within-subject case and an average accuracy equal to 80.2 % in the cross-subject case, were obtained.