search

ALIZE-Speaker-Recognition / LIA_RAL

A high-level toolkit for speaker recognition, build on top of ALIZE-Core.
http://alize.univ-avignon.fr
GNU Lesser General Public License v3.0
125 stars 27 forks source link

Please explain the expected contents of the input files to the i-vector part #35

Open zsogitbe opened 4 years ago

zsogitbe commented 4 years ago

Hello,

I have not found any info about ivNorm.ndx and Plda.ndx. The only example '02_i-vector_system_with_ALIZE3.0' contains an ivNorm.ndx file and a Plda.ndx file which are both the same:

xaaf xaag xaao
xabr xabu xaca
xacf xacs xacv
xacz xado xadz
xaei xaek xaen
...

I do not have access to the NIST data and I can not see if the 3 audio files in sequence (e.g. 'xaaf xaag xaao') are coming from the same speaker or not. I would expect that at this stage the system would need a per speaker row.

Please explain what these two files need to contain. Thank you!

Best regards, Zoltan