User Activity

  • Posted a comment on discussion Help on CMU Sphinx

    The non utf-8 characters are random as far as I can tell. Seems like corrupted memory, most likely an issue with the bindings. I cannot share the entire code base but could share small snippets. You'd better localize the problem on the server with a simple reproducible example.

  • Posted a comment on discussion Help on CMU Sphinx

    Sorry, your questions are too basic, you can answer them yourself.

  • Modified ticket #479 on CMU Sphinx

    Cross compiling to arm : error after doing make

  • Posted a comment on ticket #479 on CMU Sphinx

    Already asked at https://github.com/cmusphinx/sphinxbase/issues/49

  • Posted a comment on discussion Sphinx4 Help on CMU Sphinx

    However, I could only display the Langugae HMM if it is statically produced, correct? There are several linguists, FlatLinguist is static, others are dynamic and create states on the fly. Usually you can not display a HMM because it is a very large structure, you can do it only for very simple cases. Or, finally, am I all wrong and Java clears up the entire HMM after decoding occurs? Core linguist structure stay alive, temporary objects are released.

  • Posted a comment on discussion Help on CMU Sphinx

    It is better to write words into the grammar like "one", "thirty" and wrote the code to convert the text into actual number. This is what rasa is doing for example.

  • Posted a comment on discussion Help on CMU Sphinx

    Please suggest me know that after applying the adaptaion I am getting accuracy for the words I was applying adaptaion for but it is side affecting other sentences. It could happen. Speech recognition algorithms are statistical and changes in parameters could cause side effects. Usually the more adaptation data you use, the more accurate system is. Please let me know where I am making a mistake. I do not see any mistake here.

  • Posted a comment on discussion Sphinx4 Help on CMU Sphinx

    I'm sorry, your understanding is very rough, lattices are not created from HMM, they are constructed during search process in SearchManager. You can read it all in the code of course. If you want an overview, start with a whitepaper: https://www.cs.cmu.edu/~rsingh/homepage/papers/sphinx4_eurospeech2003.pdf and, if you need more details, with a textbook: https://www.amazon.com/Spoken-Language-Processing-Algorithm-Development/dp/0130226165

View All

Personal Data

Username:
nshmyrev
Joined:
2002-08-21 17:40:35
Location:
Moscow / Russia / MSK
Gender:
Male

Projects

  • Project Logo CMU Sphinx Speech Recognition Toolkit Last Updated:
  • Conglomerate   Last Updated:
  • FestLang   Last Updated:
  • Project Logo Kaldi   Last Updated:

Skills

  • No skills entered.

Personal Tools