| 摘要 |
PROBLEM TO BE SOLVED: To provide a method for generating a head animation.SOLUTION: A head has a mouth that moves in accordance with a voice, an input associated with lip movement and voice is divided into sequences of acoustic units, an expression characteristic for text to be inputted is selected, and the sequence of acoustic units is converted into a sequence of image vectors including a plurality of parameters that define the face of the head by using a statistical model having a parameter for describing the probability distribution that correlates the acoustic units to the image vectors. The sequence of image vectors are outputted as a video so that the mouth of the head moves and the voice associated with the inputted text is imitated accompanying the selected expression. A prescribed type parameter of each probability distribution in the selected expression is represented as a weighted sum of the same type of parameters, and a weight of expression dependency for the expression for which the conversion of the sequence of acoustic units into the sequence of image vectors is selected is searched. The weight is expression dependent, the parameters are prepared in clusters, and the weight of expression dependency is searched for each cluster.SELECTED DRAWING: Figure 2 |
