Sistema de reconhecimento de expressões faciais para classificação de emoções de usuários em sistemas computacionais

Abstract

In the software development process, interface design has gained a lot of strength in the last decades and has become one of the main tasks of this process, always seeking interfaces that achieve a human-computer interaction as "friendly" as possible. To achieve good levels of usability. It is common to record, through video, the user's interaction with the system during a usability test, an affectivity analysis among other applications. However, the use of image processing techniques in the records generated to reduce the time of the analysis is still little explored, especially when dealing with the recognition of faces and facial expressions. The present work proposes an automated system capable of receiving the inputs generated from user interaction with a computer system, such as images and videos of users during interaction with the system, and apply the face recognition technique and facial expressions. For this, the expressions were categorized in joy, disgust, anger and surprise. For the classification of the expressions a Convolutional Neural Network was used using the increased data set of the JAFFE base. Good results were achieved by reaching a 62% accuracy considering the low number of images used for training given the complexity of classifying facial expressions.