Speech Recognition project

[Eric S. Johansson]

RSousa wrote:
> Hello!! 
> 
> I'm working in a development project about voice recognition. The first
> 
> 
> step of this project, is to find real problems of used methologies. For
> 
> 
> example, in call centers, voice servers, web applications and so on. 
> 
> 
> 
> So I'd like to read some opinions to discuss about recognition problems
> 
> 
> and I sincerely thank you all. 

Not to be pedantic (but I am) which are looking for speech recognition. 
  Voice recognition looks for the uniqueness in a voice to uniquely 
identify an individual.  Speech recognition looks for the elements of 
what they are saying to translate to a textual form.

This is well trodden ground.  There are a few well-known centers of 
research for speech recognition.  Carnegie Mellon is one with the Sphinx 
Project.  So Google for speech recognition, interactive voice response 
systems (IVR) read white papers from nuance, speech works, IBM research 
to name a few.

In fact, buy yourself a copy of NaturallySpeaking (version 9 came out 
today, $200), boot up a copy of XP and play with speech recognition 
before going any further.  Once you've done this then you might be able 
to start formulating the questions you need to ask before you go any 
further.

And most of all, don't even think of trying to run speech-recognition on 
Linux.  Unless nuance had a major life change, NaturallySpeaking 9 
probably won't work under wine and you are far better off spending your 
efforts understanding the user's experience with speech recognition than 
  going on the path that may fatally distract you from your actual task.


Focus on the job at hand, learn by using about practical speech 
recognition and then start mucking with the toolkits from Carnegie 
Mellon once you understand how it fails and what are the symptoms of 
failure.  I believe this would be the fastest path to success based on 
10 years of living with speech recognition as my primary input device 
for computers.

---eric

> 
> 
> 
> 
> 
> Many thanks 
> 
> 
> 
> 
> 
> Sorry for my english. It is a little rusty!
> 
>