Garsų trukmės modeliavimas naudojant klasifikavimo ir regresijos medžius

Abstract

Currently, intelligibility of the best TTS systems is extremely good, and certainly good enough for many real applications. However, it definitely lacks naturalness. It is commonly assumed that lack of natural prosody is the main reason for this. It is generally accepted that, next to intonation, timing plays a crucial role for encoding and decoding speech.[...]