This thesis covers facial expression recognition from camera images to improve human-machine communication. A three-dimensional face model, which is fitted to the image, is used for this task. Automated facial expression recognition systems are confronted with two characteristic challenges: In contrast to artificial objects, human faces differ a lot with respect to appearance and shape. Furthermore, because obtaining natural training data is difficult, most databases provide only acted facial expressions. These challenges are tackled separately: A novel preprocessing method highlights specific facial components like eyebrows or lips. Moreover, several databases are combined instead of relying on only one database.
«This thesis covers facial expression recognition from camera images to improve human-machine communication. A three-dimensional face model, which is fitted to the image, is used for this task. Automated facial expression recognition systems are confronted with two characteristic challenges: In contrast to artificial objects, human faces differ a lot with respect to appearance and shape. Furthermore, because obtaining natural training data is difficult, most databases provide only acted facial ex...
»