F{sub}0 contour generation based on the generation process model and statistical methods considering syntactic structures

摘要

A model-constrained corpus-based synthesis strategy has been developed for fundamental frequency (F{sub}0) contours of Japanese sentences. In the training phase, the relationship between linguistic factors and the command values (amplitudes and locations) of F{sub}0 contour generation process model is learned for a prediction module. In the synthesis phase, the prediction module is used to generate command values of the model. Syntactic structure information automatically obtainable from text is added to input parameters of the statistical methods. Experimental results showed that the new parameter was effective for improving especially phrase component estimation, leading to a reduction in mean square errors of F{sub}0 patterns.