AAM based continuous facial expression recognition for face image sequences
Sebastian Hommel, Uwe Handmann · 2011
In this paper a method of automatic real-time capable continual facial expression recognition becomes described and compared with a similarly classification. Whereas the classification maps each image into one of the 7 basic emotions (neutral, happy, sad, disgust, surprise, fear, anger) and the regression maps each image into an one-dimensional emotion space. Both methods, the continual recognition and the classification based on Active Appearance Models (AAMs) and Support Vector Machines (SVMs). To reduce the influence of individual features an Individual Mean Face (MF) is estimated over time. The emotion regression will be used in service robotic by human-robotic interaction to analyze the continual change of humans emotional state to get a feedback for a gentler, more natural and adaptive dialog. This is needed for more acceptance and usability of service or health-care robotic in the household environment.