Corpus-Based Techniques in the AtT Nextgen Synthesis System

摘要

The AT&T text-to-speech (TTS) synthesis system has been used as a framework for experimetning with a perceptually-guided data-driven approach to speech synthesis, with primary focus on data-driven elements in the "back end". Statistical training techniques applied to a large corpus are used to make decisions about predicted speech events and selected speech inventory units. Our recent advances in automatic phonetic and prosodic labeling and a new faster harmonic plus noise model (HNM) and unit preselection implementations have significantly improved TTS quality and speeded up both development time and runtime.