Can I use CMUSphinx4 to decode a particular speech into phonemes and get those phonemes into further implementations?
-
1No, it's not supported. An exact duplicate question is http://stackoverflow.com/questions/12048163/cmu-sphinx4-phoneme-dictation – Nikolay Shmyrev Jan 26 '13 at 11:07
-
Then could you please tell me how the sphinx4 recognize the speech (the methodology). – Sameera Jan 26 '13 at 13:43
-
1Sure, you can check the tutorial http://cmusphinx.sourceforge.net/wiki/tutorialconcepts Speech recognizers do not look for phonemes, because phonemes pronuncation depends on context. Instead, speech recognizers look for words as a whole. – Nikolay Shmyrev Jan 26 '13 at 14:18
2 Answers
If you're using sphinx4 (beta 6), there is a method within the framework called getBestPronunciationResult()
"The method is used when the application wants the phonemes on the best final path. Note that words may have more than one pronunciation, so this is not equivalent to the word path e.g. one[HH,W,AH,N] to[T,UW] three[TH,R,IY]
Returns: the String of words and associated phonemes on the best path"
Location: src/sphinx4/edu/cmu/sphinx/result/Result.java
Token.java is the class which does the actual work.
Hope this helps.

- 43
- 6
Found this sample code that would take in a recognition result object and dump the phonemes :http://cmusphinx.sourceforge.net/wiki/sphinx4:unittimestamps
I found this paper to be very useful in understanding Sphinx4 : http://cmusphinx.sourceforge.net/sphinx4/doc/Sphinx4Whitepaper.pdf
Once I understood the architecture, I went through the list of packages found on the javadoc page.
Hope that helps. Good luck !
-
Note that link-only answers are discouraged, SO answers should be the end-point of a search for a solution (vs. yet another stopover of references, which tend to get stale over time). Please consider adding a stand-alone synopsis here, keeping the link as a reference. – kleopatra Jul 30 '13 at 22:36