Update 6/6 2012: A grievous mistake in sample generation was found and it is possible that the tests must be run again with new samples. The samples here have also been taken offline for the same reason.

Confidence-Weighted Adaptation of HMM-based Speech Synthesis Using Noisy Data

Reima Karhila, Ulpu Remes and Mikko Kurimo
Speaker adaptation with 100 sentences, as used in the listening tests for Interspeech 2012.


Training sentence samples
Synthesis samples


Clean speech
F:
M: 		F:
M:


Babble-corrupted speech
F:
M:


Factory-corrupted speech
F:
M:


Babble-corrupted speech
with noise suppression
F:
M:


Factory-corrupted speech
with noise suppression
F:
M:



Back to Aalto ICS speech group demo page