ICC   25427
INSTITUTO DE INVESTIGACION EN CIENCIAS DE LA COMPUTACION
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
A simple geometric-based descriptor for facial expression recognition
Autor/es:
ACEVEDO, D.; NEGRI, P.; GÓMEZ FERNANDEZ, F.; MEJAIL, M.; BUEMI, M. E.
Lugar:
Washington D C
Reunión:
Workshop; First International Workshop on Adaptive Shot Learning for Gesture Understanding and Production; 2017
Resumen:
Abstract? The identification of facial expressions with hu-man emotions plays a key role in non-verbal human com-munication and has applications in several areas. In thiswork, we propose a descriptor based on areas and angles oftriangles formed by the landmarks from face images. We testthese descriptors for facial expression recognition by means oftwo different approaches. One is a dynamic approach whererecognition is performed by a Conditional Random Field (CRF)classifier. The other approach is an adaptation of the k-NearestNeighbors classifier called Citation-kNN in which the trainingexamples come in the form of sets of feature vectors. An analysisof the most discriminative landmarks for the CRF approachis presented. We compare both methodologies, analyse theirsimilarities and differences. Comparisons with other state-of-the-art techniques on the CK+ dataset are shown. Even thoughboth methodologies are different from each other, the descriptorremains robust and precise in the recognition of expressions.