Time-Varying Spectral Analysis of Single-Channel EEG: Application in Affective Protocol

Abstract

Neural correlates of emotions have been widely investigated using noninvasive sensor modalities. These approaches are often characterized by a low level of usability and are not practical for real-life situations. The aim of this study is to show that a single electroencephalography (EEG) electrode placed in the central region of the scalp is able to discriminate emotionally characterized events with respect to a baseline period. Emotional changes were induced using an imagery approach based on the recall of autobiographical events characterized by four basic emotions: “Happiness”, “Fear”, “Anger”, and “Sadness”. Data from 17 normal subjects were recorded at the Cz position according to the International 10–20 system. After preprocessing and artifact detection phases, raw signals were analyzed through a time-variant adaptive autoregressive model to extract EEG characteristic spectral components. Five frequency bands, i.e., the classical EEG rhythms, were considered, namely the delta band (δ) (1–4 Hz), the theta band (θ) (4–6 Hz), the alpha band (α) (6–12 Hz), the beta band (β) (12–30 Hz), and the gamma band (γ) (30–50 Hz). The relative powers of the EEG rhythms were used as features to compare the experimental conditions. Our results show statistically significant differences when comparing the power content in the gamma band of baseline events versus emotionally characterized events. Particularly, a significant increase in gamma band relative power was found in 3 out of 4 emotionally characterized events (“Happiness”, “Sadness”, and “Anger”). In agreement with previous studies, our findings confirm the presence of a possible correlation between broader high-frequency cortical activation and affective processing of the brain. The present study shows that a single EEG electrode could potentially be used for the assessment of the emotional state with a minimally invasive setup.