Skip to content

Instantly share code, notes, and snippets.

@MisaOgura

MisaOgura/gmm_suggestions.md

Last active March 21, 2019 13:36
Show Gist options
  • Save MisaOgura/dba04df185fcb09c6839d79ac6db4376 to your computer and use it in GitHub Desktop.
Save MisaOgura/dba04df185fcb09c6839d79ac6db4376 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
gmm_suggestions.md

Things we can try with GMM (suggested during a conversation with Jana)

  • Increase the training iteration to few hundreds

    • Possible stopping criteria: min variance (keep an eye out for collapsing variance)

  • Increase the number of Gaussian curves within each model

    • Currently init 32, diag 30 and full 20
    • Try init 258, diag 128 and full 64?

  • Increase the size of MFCC

  • Delta is useful, Delta-Delta not so much

  • Try Viterbi decoding for more intelligent decision smoothing?

  • Change window size?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment