This would work if you only care that you get very rough phonetic
spellings as Don implied. If you think about it humans cannot do any
better. I know personally - I have to spell my name all the time.
Perhaps your app could ask them to spell their name, which actually has a
shot at
There are solutions ranging from free to many thousands of dollars, with
effectiveness ranging from nearly worthless to almost pretty good.
A lot depends on your application.
The most successful application would match an utterance from a known
speaker to a known list of a couple dozen
I wouldn't have a database to compare names to, each one would essentially be
unique and unknown. It's sounding like this idea may be not possible...What
high end options are available? I read about lumenvox, but I believe that
compares to a known list of names (such as a directory, or Yes No,
Hi Alfred,
There is a research project by Carnegie Mellon University (CMU) on a very
versatile Speech Recognition Software. Its Sphinx
http://cmusphinx.sourceforge.net/html/cmusphinx.php . This application is in
raw state and the Version 2 of sphinx could be integrated with Asterisk.
Festival
Thanks. This has been useful, I've been toying with Sphinx4 but haven't gotten
very far with it yet. Took a while to get my environment set up to compile all
the java sources.. should have downloaded the binary distro...:-)
From: Kurian Thayil
