Man vs. Machine in Speaker Recognition Part 1

In, forensic speaker verification, we aim to detect whether
same speaker is speaking in the incriminating recording and
in the so called test recording. The verification task is
generally performed either by professional phonetician
alone or assisted by an automatic speaker verifications system.
NIST Speaker Recognition Evaluation 2010 contains a Human
Assisted Speaker Recognition (HASR) sub-challenge. It gives
us a new opportunity to study human vs. machine performance
in speaker verification. We will study how to use pools of
so called naive listeners and combine results of those with
the state-of-the-art automatic verification system.

Category:

Education

Tags:

• IEEE