METHOD AND APPARATUS FOR TRAINING DIFFERENCE PROSODY ADAPTATION MODEL, METHOD AND APPARATUS FOR GENERATING DIFFERENCE PROSODY ADAPTATION MODEL, METHOD AND APPARATUS FOR PROSODY PREDICTION, METHOD AND APPARATUS FOR SPEECH SYNTHESIS

摘要

PROBLEM TO BE SOLVED: To provide a method and apparatus for training a defference prosody adaptation model capable of easily generating an accurate and stable defference prosody adaptation model. SOLUTION: An initial parameter estimation model, which includes a plurality of attributes relating to defference prosody estimation and at least part of a plurality of attribute combinations obtained by combining the plurality of attributes respectively as items, for the respective parameters of a defference prosody vector expressed by a prosodic duration and a coefficient of an FO orthogonal polynomial. An item with the lowest importance calculated on the respective items is deleted from the parameter estimation model. The items with low importance are deleted from the parameter estimation model until the parameter estimation model formed of the remaining items becomes an optimum model. The defference prosody adaptation model including the defference prosody vector and the parameter estimation model of the respective parameters determined as the optimum model is obtained. COPYRIGHT: (C)2009,JPO&INPIT