Personal tools
You are here: Home → Bibliography → Variable-length acoustic units inference for text-to-speech synthesis
Log in


Forgot your password?
 

previous entry previous entry    reference list reference list    next entry next entry

Boëffard, O. (2001). Variable-length acoustic units inference for text-to-speech synthesis. Proceedings of Eurospeech 2001. Aalborg, Denmark.
search for this reference on Google Scholar
search for this reference on Microsoft Live
 

imported comment

Posted by Marc Schröder at Monday, 27 June 2005

Algorithm. Apply the EM algorithm in order to find the best subset of n-multigrams (= 1-grams, 2-grams, ..., n-grams) representative of a big corpus. Very mathematical. Based solely on phonetic symbols for describing the ``atoms'', not F0, position, etc.
 
Powered by Plone

Portal usage statistics