Clock work hierarchy variational encoder

摘要

A method (400) for representing the intended prosody in the synthetic tone (152) includes the step of receiving a text utterance (320) having at least one word (250) and selecting an utterance burial (260) for text utterance.Each word in the text utterance has at least one syllable (240) and each syllable has at least one phoneme (230).The utterance emulation represents the intended prosody.For each syllableUsing the selected utterance embeddingThe method isCorresponding prosodic syllable embedding (245) for syllablesBy encoding the language features (232) of each phoneme of syllablesPredicts the duration of syllables andPredicting pitch contours of syllables based on prediction duration for syllables andBased on prediction duration for syllablesIncluding generating a plurality of fixed length predictive pitch frames 280.Each fixed length prediction pitch frame represents part of the predicted pitch contour of syllables.