Abstract: |
The goal of the dissertation is to identify a human emotion through analysis of facial expressions in a short video sequence. As first steps, the video clip is split into frames and the face is localized. Within each frame a characteristic set of features is extracted, which contain mouth, eyes, and brows positions. We trace the changes of the feature set in time - the so called facial dynamics in time (i.e. along the whole video sequence), and assume that there are trends in these changes that characterize a concrete human emotion (of happiness, sadness, anger, etc.) |