Status report - Voice recognition in khotkeys.

Olivier Goffart ogoffart at gmail.com
Wed Aug 3 16:11:48 CEST 2005


Hello

This is my first status report.
I had not done status report because i was not coding a lot last month because 
of vacations.


My project is speech recognition in khotkeys

I have firstly worked on a algorithm to do speech recognition myself, without 
using other's people code. Just by using knowledge I have
I wanted to do it alone for fun, to see if i was able to do it.
It works a bit,   but i can't say it is really usable.
I wanted to implement myself the FFT (fourrier transform), it took me some 
time, but i did it successfully

Anyway, I already integreted this engine in khotkeys.
The khotkeys coding style is very different from what I am used to.

It is actually possible to set have a voice trigger.

If you want to test it, you need to checkout kdebase svn (branch KDE 3.5)
and then, in  kdebase/khotkeys
svn switch 
https://svn.kde.org/home/kde/branches/work/soc-khotkeys-speechrecognition

There should be at least 3 different voice command in order to get results.
But results aren't so good.

Currently, to trigger the voice command,  press F11 , talk , press F12.   In 
the future, the shortcut will be a single (configurable) key.  I still don't 
know what the default key should be.


Now, I'll finish the interface, which is far from finished,  and make the 
speech recognition algorithm better.
I have already searched on google for documents, and found some really 
interesting documents that explain how speech recognition works.
And i will try to make the speech recognition 


so TODO:
- Improve the speech recognition algorithm
- Use a configurable single key as trigger
- Make sure that new voice command are different enough of existing ones.
- Finish the kcontrol interface.


Sorry for my bad English.
Olivier


More information about the Kde-soc mailing list