Generation and Evaluation of Prosodically Varied Synthesized Speech
Harm Laméris Annual PhD Progress Seminar
Tid: Fr 2023-10-27 kl 15.00 - 16.00
Plats: Fantum (Lindstedtsvägen 24, floor 5, room no. 522)
Språk: English
Kontakt:
Respondent: Harm Lameris
Abstract
During this 50% seminar I will discuss the work I have published at ICPhS, Interspeech, and SSW over the last 12 months. My work focusses on using spontaneous speech data for TTS, through which we are able to synthesize naturalistic speech phenomena, such as creaky voice. Another aspect is the evaluation of the prosodically varied speech we are able to produce, using methods such a in-context evaluation.