抄録
In this paper, a human centered approach for interactive face drawing is presented. Human users of the system have the possibility to draw and retouch faces from linguistic terms. The system learns the meaning of the terms interactively from the retouching process.
The system has an average face of each race, generated using anthropometric measurements of the face and the head. An average face corresponding to the race of the target face is displayed to a user as an initial face. A user describes each facial feature of the desired face using linguistic descriptors and the feature changes according to the user's description. Depending on the race of a user and the gender of the face to be drawn, the relation between the size of each facial feature and the set of the linguistic descriptors used to describe it is obtained by conceptual fuzzy sets. These fuzzy sets represent the initial meaning of the linguistic terms.
Once the face is obtained, a user can retouch the drawn face. The user uses a combination of an adverb and a linguistic descriptor to retouch some features of the drawn face until the desired face is obtained. For the current user, the system interactively learns his/her meaning of the descriptors using the proposed Online Learning of Gaussian Fuzzy Sets algorithm based on the Maximum likelihood estimation (MLE) for Gaussian mixture. The system then adjusts the drawn face according to the new meaning of the descriptors.
