Combined ASR-TTS systems and prosody prediction

[Evelyn W]

Has any research been done on transferring (or using as context) recognised prosody in combined ASR-TTS systems?

I’m thinking specifically about assistive voice technology, where an ASR component could extract prosodic features from a conversation partner and use them to improve prosody prediction in the synthesis component.

I guess this could also apply to voice assistants. I had a look but couldn’t find anything related to this topic

[Evelyn W]

I found this one paper from Interspeech 2020 which might be what I’m looking for https://arxiv.org/abs/2009.01475

[Author]

Posts