-
Notifications
You must be signed in to change notification settings - Fork 0
Voice Authentication
AishwaryaKalloli edited this page Feb 27, 2019
·
3 revisions
- Identify speaker's mouth and throat shape for parameters.
- They are unaffected by mimicry or change in mood.
- User's voice print is created and compared with everytime.
- Where acan voice authentication be used SANS Institute
- Main thing to keep in mind: Recorded voice should not be valid. Person should be there, how to do that?
- Can be used to catch mimicry(by police force etc).
- Where all voice authentication can be used
Two ways for KS and SR: verification, identification
One time: verification(supervised-classification)
Continue: identification(unsupervised)
Dataset for SR:LibriSpeech
devclean
40 different speakers