speaker-recognition icon indicating copy to clipboard operation
speaker-recognition copied to clipboard

Published 20 hours ago • verified publisher icon ppwwyyxx

Defining reject_threshold for reject prediction.

Open rhythmize opened this issue 7 years ago • 0 comments

After successfully training both UBM model and user specific GMM models. The difference between the ubm_score and the maximum gmm_score shows large variance for both trained and untrained users, so I'm not able to define reject_threshold for the system. Is there a way to resolve this?

P.S. : Is this the reason you updated UBM_MODEL_FILE to None in 'src/gui/interface.py'?

rhythmize avatar Oct 02 '17 05:10 rhythmize