Speaker age estimation for elderly speech recognition in European Portuguese

Phone-like acoustic models (AMs) used in large-vocabulary automatic speech recognition (ASR) systems are usually trained with speech collected from young adult speakers. Using such models, ASR performance may decrease by about 10% absolute when transcribing elderly speech. Ageing is known to alter s...

ver descrição completa

Detalhes bibliográficos
Autor principal: Dias, J. (author)
Outros Autores: Pellegrini, T (author), Hedayati, V. (author), Trancoso, I. (author), Hämäläinen, A. (author)
Formato: conferenceObject
Idioma:eng
Publicado em: 2022
Assuntos:
Automatic speech recognition
Elderly speech
Automatic age estimation
I-vector extraction
Texto completo:http://hdl.handle.net/10071/25451
País:Portugal
Oai:oai:repositorio.iscte-iul.pt:10071/25451

Texto Completo

http://hdl.handle.net/10071/25451

Registos relacionados