Menu

Help training UBM with Lium Diarization

Help
chen
2017-10-26
2018-01-25
  • chen

    chen - 2017-10-26

    Hi,

    I would like to train a UBM GMM then speaker specific GMMs for speaker diarization using the LIUM Diarization toolkit. I have been following the documentation on the LIUM website (http://www-lium.univ-lemans.fr/diarization/doku.php/gaussian_gmm_training) but am having trouble understanding how to point to multiple feature files in the training process.

    I am using the following command:
    java -Xmx2048m -cp ../lium_spkdiarization-8.4.1.jar \ fr.lium.spkDiarization.programs.MTrainInit --nbComp=8 --kind=DIAG --emInitMethod=split_all --emCtrl=1,5,0.05 \ --sInputMask="./%s.seg" \ --fInputMask="./%s.mfcc" --fInputDesc="sphinx,1:1:0:0:0:0,13,0:0:0:0"
    --tOutputMask="./%s.init.gmms" "test"

    The howto page describes how the segmentation file (.seg) can contain segmentations of multiple files. But how do I specify the different feature files (.mfcc) corresponding to the different files named in the segmentation file?

    Thank you

     
    • Nickolay V. Shmyrev

      Use kaldi diarization

       

Log in to post a comment.

Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.