The amount of movement learning device of the fundamental frequency for speaker adaptation, the fundamental frequency generating device, the amount of movement learning method, the amount of movement learning program and the fundamental frequency generation method

摘要

Based on only a small amount of training data, to provide a technique that can accurately reproduce the characteristics of the fundamental frequency of speech of the target speaker. Learning device to learn the amount of movement of the target F pattern of the target speaker to the original F0 pattern used as the reference, the mountains, and the original F0 pattern corresponding to the training text, and the target F0 pattern corresponding to the learning for the same text associates the troughs and valleys so as to correspond to, for each point in the F0 pattern objectives, the frequency axis and the time axis direction from the point of view the results of the mapping, the corresponding original F0 pattern to obtain the movement amount, the input feature value, and to learn a decision tree as the output feature value movement amount calculated language information is a result of analysis of training text.