Reima Karhila, D.R. Sanand, Mikko Kurimo and Peter Smit
Speaker adaptation with 10 sentences, as used in the listening tests for ICASSP 2012.
(See the paper in IEEExplore)
|
40 child speakers 60 sentences each |
|||
VTLN normalisation | ||||
|
CSMAPLR group adaptation | CSMAPLR group adaptation | ||
Target speaker |
||||
CSMAPLR speaker adaptation with 10 sentences | CSMAPLR speaker adaptation with 10 sentences | VTLN + CSMAPLR speaker adaptation with 10 sentences | CSMAPLR speaker adaptation with 10 sentences | |
Adapted from adult voice |
Stack adapted voice |
Stack adapted voice with VTLN |
Adapted from child voice |