Non-linear active model for mouth inner and outer contours detection

Mouth segmentation is an important issue which applies in many multimedia applications as speech reading, face synthesis, recognition or audiovisual communication. Our goal is to have a robust and efficient detection of lips contour. In this paper, we focus on the detection of the inner mouth contour which is a difficult task due to the non-linear appearance variations. We propose a method based on a statistical model of shape with local appearance gaussian descriptors. Our hypothesis is that the response of the local descriptors can be predicted from the shape. This prediction is achieved by a non-linear neural network. We tested this hypothesis with a single speaker task and compared the results with previous methods. Then this approach is generalized to take care of the intra person appearance variability in a multi-speaker task.