Abstract
Frontal alpha asymmetry (FAA), a pattern of brain activity that reflects the difference in alpha wave power between the left and right frontal areas of the brain, is considered a stable marker for an individual's tendency to experience either more approach-related or withdrawal-related emotions. On the other hand, electrodermal activity (EDA) measures arousal by tracking changes in skin sweat, which are controlled by the sympathetic nervous system. This study explores the interrelation between EDA features, obtained from time and frequency domains, with FAA by means of the mutual information. Multiple cognitive tasks such as EAT, ship search, PVT and N-Back were analyzed in 10 participants in intervals of two hours over 24 h (12 trials), in which they had to face sleep deprivation conditions. The most informative EDA features about FAA, were used to identify the two main clusters associated to high and low FAA values through the hierarchical agglomerative clustering approach. Once data is labeled, a supervised classifier based on support vector machines (SVMs) is used to identify positive and negative emotional states by using a rigorous one-trial out cross-validation scheme. Results show consistent performance within tasks and trials, achieving accuracy values over 80% on average, giving an important insight about the use of EDA signal as an alternative to the more complex FAA measurement for tracking positive or negative emotional states.