The recognition of emotions in facial expressions by computer is an emerging and challenging problem to be explored in various areas. In the context of human and behavioral cognition, understand is seeked if there is consensus between the forms of expression to which individuals externalize, whether they come from biological or social factors. In the computational area, understanding computationally what humans are expressing through captured images or videos opens up a range of possibilities to be explored. Computational resources can be directed on demand, according to what the person needs through the interpretation of their expression, their utility is associated with the personalized delivery of content in the area of games, marketing and business intelligence, learning, tutoring systems, studies related to neuroscience, psychology and human cognition and behavioral, among many others. Developing algorithms that have the ability to interpret and read facial expressions performed by humans, is a tool of relevant importance to approximate the interaction performed by human-computer. Thus, this research project aims to study the problem in question and to present computational solutions to the problem of emotion recognition in facial expressions. At first, an extensive search for computational mechanisms for the solution of the proposed problem will be performed, as well as tests of the found solutions and development of the computational solution. At the same time, the objective is the elaboration of a proper dataset of faces captured from different people, associated with the meaning of the respective facial expression presented, be it fear, anger, satisfaction, neutral and others. With the dataset of images, artificial intelligence will be utilized, more specifically to verify the use of convolutionalneural networks to recognize the presented expression. This project aims to prepare an article to be submitted in a event or qualified journal, as well as the technical report and dataset to be made available to the scientific community.

Project Registered in SIGPEX UFSC 201902487.