[Sugar-devel] Speech Recognition

[Tony Anderson]

You are correct. The use-case is quite different from that of most 
researchers in the field. The intial goal was continuous recognition of 
speech such as dictating a document. The current popularity of the 
spoken command is much more realistic (limited vocabulary, predictable 
syntax). However, in both cases the speech recognizer tries to 
understand the speaker's diction (hence the training).

What is probably needed is to train a recognizer with a range of native 
speakers of the language and then see whether the learner's speech can 
be recognized, but that the learner will learn to speak recognizably.

I don't have any experience as a developer but have been following this 
for a long time. The CMU project has been going on for years so it a 
good choice for stability and has a university research team behind it.

I have tried matching the oscilloscope sound, but did not find it 
helpful (French réfrigérateur proved impossible).

Tony

On 04/28/2016 09:00 PM, Justin Overton wrote:
> Tony,
>
> Do you have experience with this? So far as I can tell the "compare 
> sound patterns" part gets complicated.
>
> This use-case differs from most of the research I've found regarding 
> speech recognition in that normally a recognizer will try to determine 
> the words of a given audio-snippet. This use-case is simply to test if 
> a given audio-snippet is a specific word or phrase. I imagine the 
> latter is less complicated, but I haven't found any out-of-the-box 
> solutions for it.
>
>
> ---- On Wed, 27 Apr 2016 18:27:15 -0500 *Tony Anderson 
> <tony_anderson at usa.net>*wrote ----
>
>     The application of speech recognition for Sugar is to enable
>     learners to pronounce words or sentences in English (or other
>     second language) and have the speech recognition affirm by
>     recognition that the sounds are sufficiently accurate. Currently
>     this is often done by having the learner compare sound patterns
>     (oscilloscope style).
>
>     As a simple example, suppose the student is presented an
>     arithmetic problem (e.g. by KA Lite). Give the student the ability
>     to say the result instead of having to enter it from a keyboard.
>
>     Tony
>
>
>     On 04/28/2016 01:56 AM, Justin Overton wrote:
>
>
>     _______________________________________________
>     Sugar-devel mailing list
>     Sugar-devel at lists.sugarlabs.org
>     <mailto:Sugar-devel at lists.sugarlabs.org>
>     http://lists.sugarlabs.org/listinfo/sugar-devel
>
>         The application I am working on, and targeting across multiple
>         platforms is HTML5 + JavaScript for ease of portability. Right
>         now I have a very naive Speech Recognition system in place,
>         but I plan to use CMU's PocketSphinx (particularly the
>         Pocketsphinx.js port).
>
>         Has anyone else used any speech recognition in Sugar (or web)?
>         Would anyone be interested in working with me on creating a
>         speech recognition library that can be used throughout Sugar
>         (native and web-based)?
>
>
>
>         _______________________________________________
>         Sugar-devel mailing list
>         Sugar-devel at lists.sugarlabs.org
>         <mailto:Sugar-devel at lists.sugarlabs.org>
>         http://lists.sugarlabs.org/listinfo/sugar-devel
>
>
>
>
> _______________________________________________
> Sugar-devel mailing list
> Sugar-devel at lists.sugarlabs.org
> http://lists.sugarlabs.org/listinfo/sugar-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sugarlabs.org/archive/sugar-devel/attachments/20160429/c9d83d17/attachment.html>