[Olympus developers 96]: Re: Endpoint of utterance..

[Thomas Harris]

Hi Νίκος,

So I just created a short enpointing article that outlines the basic
mechanism. Antoine hopefully will correct any of my misconceptions as he's
the author of the mechanism.

http://wiki.speech.cs.cmu.edu/olympus/index.php/Endpointing

Thanks,
-Thomas

On Mon, Jan 26, 2009 at 7:44 AM, Νίκος Κατσαούνος <nkatsaounos at sch.gr>wrote:

>  Hi again!
>
> I set-up an Olympus microphone-based application that uses PocketSphinx for
> ASR.
> My problem is that when I make my request, in case no recognition
> result/concept is produced (for any reason, eg fault of recognition engine,
> no definition in lexicon, no definition in grammar etc), the application
> still waits for an input utterance.
> The behaviour that I want my application to have, is to *proceed with a
> corresponding prompt either there is a valid recognition result or not.*
> So, my question is, *how can I define an "endpoint" to user utterance that
> correspond to the duration of final pause/silense of user utterance (eg
> 1000msec)* ; So, when a user speaks to the microphone, after eg 1000 msec
> pause of the end of its utterance, recording process is stopped and DM
> continues, regardless of having or not recognition/concept result.
> I tried out to find something in Spinxvad component, but I didn't find the
> way I could implement/configure the above..
>
> Thanks in advance
>
> Nikos Katsaounos
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.srv.cs.cmu.edu/pipermail/olympus-developers/attachments/20090127/7db1762a/attachment.html