[Japanese | Thesis | Researches in Minoh Lab | Minoh Lab]
Facial expression mapping realizes the mapping from a facial expression on the real face to the same facial expression on the artificial face. A new method is proposed to estimate and realize the facial expression mapping in accordance with the preference of the user in distance communication environment. The facial expression mapping needs not only to reflect the preference of the user but also to obtain the precision defined based on the facial expression category information of the user. The proposed method approximates the mapping function fitting the preference of the user using the positive examples of facial expression acquiring from the user interaction, RBF (radial basis functions) network is used for realizing this process. And also the necessary facial expression categories and the required precision of the user are extracted from the positive examples and negative examples acquired from the user interaction. We also refine the RBF network based on the extracted category information so that the facial expression mapping is able to realize the required precision. Finally the simulation and experimental result show that the mapping result becomes to fit the preference of the user with the incrementally obtained positive examples of facial expression, and the mapping precision is obtained by refining the RBF network based on the user's category information.