Towards a Vowel Formant Based Quality Metric for Text-to-Speech Systems: Measuring Monophthong Naturalness
15.06.2022·,,,,,
Sven Albrecht
Rewa Tamboli
Stefan Taubert
Maximilian Eibl
Günter Daniel Rey
Josef Schmied
Abstract
This presentations presents our proposed vowel fromant based quality metric for text-to-speech systems. We describe our automated vowel formant measuring pipeline and show results for our TTS system trained on the LJSpeech dataset.
Date
15.06.2022 — 17.06.2022
Event
Location
TU Chemnitz, Chemnitz, Germany