Hello World Decoder QuickStart Guide Pt 3

From Sphinx

  • Unzip
  • Modify cfg file to use new acoustic models. Delete the -hmm param, and add 
-mdef /path/to/wsj/model_architecture/wsj_all_cont_3no_8000.mdef
-mean /path/to/wsj/model_parameters/wsj_all_cont_3no_8000_16.cd/means
-var  /path/to/wsj/model_parameters/wsj_all_cont_3no_8000_16.cd/variances
-mixw /path/to/wsj/model_parameters/wsj_all_cont_3no_8000_16.cd/mixture_weights
-tmat /path/to/wsj/model_parameters/wsj_all_cont_3no_8000_16.cd/transition_matrices
  • Output 
FWDVIT: (yellowaudio)
FWDVIT: (pleaseaudio) 
FWDVIT: (helloaudio)

NOTHING MATCHED!

Extra credit to anyone who can figure out why nothing is matched when using this acoustic model.

Retrieved from "http://sphinx.subwiki.com/sphinx/index.php/Hello_World_Decoder_QuickStart_Guide_Pt_3"
related