Landmark-based Facial Expression Parametrization for Sign Languages Avatar Animation

Resumo

Facial expressions and associates emotions important factors defining the message tone and context information in both spoken and sign language communication. Sign language virtual systems use structural models that define control values specifying body configurations in the animation process, where facial parameters are generally relegated to simple templates or completely neglected. In this work, a facial expression parametrization for avatars is proposed through an procedure that aims to identify the most relevant facial landmarks and emotions in the context of sign languages, in order to enhance automatic sign synthesis systems. An analysis of the influence of landmarks on a geometric mesh based on MPEG-4 model of human face is performed in this research, aiming to identify the principal components and their relationships, in order to allow further optimization of the animation process, supporting faster and lighter avatar animation.