Regresar

Experimental Analysis Using Action Units as Feature Descriptor for Emotion in People with down Syndrome

Abstract:

This article presents an analysis based on Action Units of the facial expressions of four basic emotions in people with Down Syndrome, such as: happiness, sadness, anger and surprise, taking as a reference the Facial Action Coding System, proposed by Paul Ekman, which is based on the universal study of the movements of the muscles of the face, used in some research in people with typical development. For the present study, the action units represent the feature extraction phase of a dataset of images of people with Down syndrome, extracted from the open access website. The tool used to obtain the features of the microexpressions for the action units was Open Face 2.0, which has open source for the development of research topics. Additionally, the statistics that most closely approximated the activation of the Action Units were obtained with their respective equations, using the probability density function, evaluated by the Kullback Leibler Divergence. The statistics obtained in this article made it possible to identify the activation of action units of some emotions that are not found in the literature, such as: AU20 in happiness, sadness and anger; AU15 and AU9 in anger that will be the basis for projecting new research on algorithms for the recognition of emotions in people with Down syndrome through the face.