RACHED, T. S.; http://lattes.cnpq.br/6659783661235661; RACHED, Taciana Saad.
Abstract:
Human emotions are fundamental to the way in which individuals interact with each
other, as well as with machines. Unlike the human-human interaction, where a human
understands another's emotional behavior, in the human-machine interaction, the last one
is unable of understand the feelings of a person, which makes hard the interaction process
among them. In this scenario, affective computing is a recent research area aiming at the
recognition of human emotions during the interaction with a machine, making it simpler,
easier and enjoyable. Several data sources are used for emotion recognition, such as facial
expressions, voice signals, body language, psychological and brain signals. Brain signals
are the most reliable data source to recognize emotions because they are not susceptible to
false simulations and ambiguous interpretation, unlike, the facial and body expressions, as
--- — well as voice signals. Despite being an important source of data, most affective detection
systems based on brain signals discussed in the literature are context free. This fact
implies a serious problem for the recognition of emotions, since the context is crucial in
identifying them. In this work we propose a method for the detection of human emotions
based on the fusion of an individual context information with their brain signals. The
method have two components: the brain signals and context. Furthermore, we developed
three case studies to evaluate the proposed method. In the first case study, the objective
was to evaluate the component of the method related to brain signals. The EEG signals
from a database were processed and classified in emotions with algorithms defined for this
purpose. In the second case study, the objective was to evaluate the context component of
the method. The relevant variables that characterize the context were identified and their
influences have been set. The context was classified in the emotions of the participants.
Finally, the third case study, the proposed method in this thesis was evaluated considering
its two components. The brain signals and the context were classified in the emotions of
the participants.