multi-speakerID baseline

This is python implementation of multi-target speaker recognition based on i-vector feature. This is also baseline system of the first Multi-target speaker detection and identification Challenge Evaluation (MCE 2018, http://www.mce2018.org )

Dataset

You can download i-vector dataset if you register and confirmed by MCE 2018 organizer. After download, extract to data folder

System flow

Performance

If you run the code like

python mce2018_baseline_dev.py

you will see the performance on top-S and top-1 detector as below :

Dev set score using train set :
Top S detector EER is 2.00%
Top 1 detector EER is 13.41% (Total confusion error is 492)

And the code also generate example submission file with name "teamname_fixed_primary.csv" and the format are [test utterance ID],[score],[Closest blacklist speaker ID] per each files. For example

pnah_431154,0.36613864,33762391
qtxw_470243,0.39585015,60587769
....

Question

Please email to mce organizer if you have question. mce@lists.csail.mit.edu or swshon@mit.edu