Detecção de pontos em faces para identificação de emoções

Abstract

Facial expressions act as a non-verbal means of communication capable of reflecting the emotional state of the human being, and through this understanding we can verify the coherence between what is said and what is manifested. Each basic emotion transmitted by an individual has facial muscle movement, therefore, a technique called Facially Ex- pressed Emotion Labeling (FEEL) is often used by researchers, in order to assess people’s ability to understand the feelings of others through their expressions, consisting of the presentation of the image of a face on a computer screen, in the sequence that same face is presented transmitting one of the six emotions considered basic: joy, sadness, fear, anger, disgust or surprise. At the end, the participants must say which feeling was expressed in the image, but not all feelings are automatically perceived with the same ease. Therefore, there is interest in the implementation of Human-Machine Interfaces (HMI) capable of identifying the expression of an individual, helping to solve this problem. In the area of robotics, there are already robots capable of identifying and replicating facial expressions similar to those of humans with the help of Artificial Intelligence (AI). Some computational techniques such as corner detection algorithms are used to identify points of interest in images, among the best known we can mention the Harris Corner Detector, used in this work with the aim of analyzing and identifying the best parameters for identifying emotions in faces images. To perform the proposed analysis, the databases The Karolinska Directed Emotional Faces (KDEF), Japanese Female Facial Expression (JAFFE) and CK+ will be used, along with Convolutional Neural Network and Gabor filter.