Menu

Combining two acoustic model

Help
asr2010
2012-02-28
2012-09-22
  • asr2010

    asr2010 - 2012-02-28

    I would like to create an acoustic model (AM) that include the phonemes of two
    different languages. given that I have 2 separate AMs for these two languages,
    how can I combine the following files: means, mixture_weights,
    transition_matrices, variances? for the mdef file, I think I can write a
    script to do the job or do you think that there is another way?

     
  • Nickolay V. Shmyrev

    There is no specific tool to do that. You can write it if you understand the
    format of the model.

     
  • asr2010

    asr2010 - 2012-02-28

    ok I will proceed with the implementation of this tool.

    Is there any tool to convert the AM binary files to ascii? I need something
    like that to check the final AM after merging?

    I guess the python module can do the job but unfortunately I couldn't find
    find documentation for it.

     
  • Nickolay V. Shmyrev

    Is there any tool to convert the AM binary files to ascii? I need something
    like that to check the final AM after merging?

    it's called printp. A part of Sphinxtrain

     
  • asr2010

    asr2010 - 2012-03-02

    Thanks. I have successfully implemented a tool to combine 2 AMs

     

Log in to post a comment.