[mythtv] Video Conferencing Plugin

[Eric A. Litman]

"Henk Poley" <hpoley at dds.nl> writes:
> Op zondag 2 mei 2004 21:02, schreef Eric A. Litman:
> > Which TTS engine are you using? Festival? If you're going that far, have
> > you thought of connecting it to Sphinx
> > (http://fife.speech.cs.cmu.edu/sphinx/) to allow voice control? If
you're
> > interested, I can share code with you for tieing Sphinx-2 to SIP
streams.
>
> Hmm, seems rather ambitious. Seems more like a 'if there's time
afterwards'.

I'd take it another step to say "if there's time afterwards and nothing left
in your life's priority queue". Still, it's a genuine offer, although my
preference would be to share the code "over the wall" and not support it (as
my own priority queue is brimming).

> As far as I can tell you need to form a system of sentences that the
system
> needs to recognise and learn it to Sphinx using SphinxTrain.

Sphinx-2 is a speaker independent system, meaning it does not need to be
trained for each user; however, you will want an acoustic model appropriate
for the medium, and the closer your accoustic model matches the parameters
of the input speech (including language, codec, etc.), the higher the
recognition rate will be. Sphinx ships with a default acoustic model for
English, and I've seen other languages/environments floating around the Net.

Acoustic models do not work on senteces per se; rather, they characterize
speech sounds (phonemes) and how those sounds relate to other sounds or
change over the course of a speech sample.

You'll also need to build an appropriate language model for MythTV, but
would be - by far - the least challenging part of this exercise.

While I'll readily admit that adding speech recognition to MythTV might seem
frivolous (and likely is given the myriad other areas in which MythTV could
benefit from improvement today), others in the industry are making a bet
that voice control of the set-up device will catch on:

http://money.cnn.com/services/tickerheadlines/prn/cgth014.P1.04282004203238.29838.htm

In this example, Motorola's partnered with AgileTV to enable operators to
offer voice control to subscribers with DCT2000 set-top boxes.

-Eric