Obtaining a string of phones from speech

[petebleackley]

I've seen a lot of language recognition software on the internet, but it's all
text based. I'd like to create a speech-based language recognizer. To do that,
I need a stripped-down speech recognition engine that will convert an
utterance to a sequence of phones, which can then be fed into an AI component
which infers the most likely language. Is it possible to use Sphinx in this
way?

[Nickolay V. Shmyrev]

Yes, it's possible. See

http://cmusphinx.sourceforge.net/wiki/phonemerecognition

[petebleackley]

That article makes it look difficult to use for my application - what I would
really like is a component that, when fed a stream of audio, will feed phones
extracted from it into my language classifier software, like the following
pseudocode

Classifier = LanguageClassifier(models) # models represents the phonotactics
of various languages
Phones = PhonemeClassifier(AudioStream) # an iterable object
Client = LangugeIdentifierClient()

for phone in Phones:
Client.UseResults(Classifier.IdentifyLanguage(phone))

It looks like I'll have to hack something together out of the source code.