From: Speaker-dependent model interpolation for statistical emotional speech synthesis
Method
Angry
Happy
Sad
Overall
Interpolation
28/45
41/45
25/45
94/135
Adaptation
17/45
4/45
20/45
41/135